Agent-readable comparison

Unique AI: sourced findings

Name: Unique AI: agent-readable findings
Creator: onAI

Published by onAI. onAI also sells one of the products compared here. Findings about other products link to public sources so you can check them; statements about onAI are marked as stated by onAI.

Unique AI has sourced findings for 49 of 51 criteria, including 49 that a second agent cross-checked. "Sourced" findings passed publication validation: they are scoped, fresh and cite public resolved sources. "Agent cross-checked" findings also received a separate agent source pass, not necessarily a hands-on test. "Stated by onAI" marks the publisher's own statements about its product (evidence type vendor-statement), which cite no third-party source. Humans approve the rubric, templates and generator; they do not review every finding. No winner is declared. A missing sourced finding is not evidence that a capability is absent.

Unique AI source checks: 2026-10-21 to 2026-10-26. Website: https://www.unique.ai
Generated 2026-10-27; next source check due 2026-12-20.

Same facts as Markdown and JSON. Methodology.

Models

MOD-01 Model catalogue and providers

Judgment rule: List each available model by provider and plan, including reasoning, fast and multimodal variants. Record whether the user picks the model, the product routes automatically, or both. A provider counts only if at least one of its current models is selectable on the scoped plan.

Unique AI

Agent cross-checked Support: partial

The platform catalogue supports several providers, subject to tenant credentials, contracts and validation status.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration; version=not specified; plan=Contracted Unique AI platform and enabled modules. No public named tariff tiers were established; deployment and provider entitlements vary by MSA.; add_ons=none stated; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

Exact contracted/activated models, automatic cross-model routing and a complete image/audio/file matrix were not established. Catalogue dates include future release identifiers; dated additions and tenant eligibility take precedence over a release number.

providers Which model providers are available? [partial]: Catalogue includes Azure OpenAI/OpenAI, Anthropic, Google Gemini/Vertex, Together AI and xAI. Provider use must be agreed with each client; a catalogue row is not proof every tenant has it. Sources: unique-ai:unique-ai:models.
current_models Which current models are selectable on this plan? [partial]: Release 2026.42, hosted 14 October, explicitly adds Azure GPT-6.1 Sol and Claude Opus/Sonnet 5.5 through Anthropic/Vertex. Availability overview also lists Gemini3.8 Flash; provider/tenant credentials and validation differ. Sources: unique-ai:unique-ai:release42, unique-ai:unique-ai:models.
user_choice Can users pick the model per chat? [partial]: Space admins set enabled models; release notes describe chat model switching and API per-message model selection under the same rules. Conduct choices depend on administrator configuration. Sources: unique-ai:unique-ai:release42, unique-ai:unique-ai:agent, unique-ai:unique-ai:conduct.
auto_routing Does the product route between models automatically, and is the chosen model shown? [unknown]: Does the product route between models automatically, and is the chosen model shown? Unknown. Exact contracted/activated models, automatic cross-model routing and a complete image/audio/file matrix were not established. Catalogue dates include future release identifiers; dated additions and tenant eligibility take precedence over a release number.
multimodal_input Which models accept images, audio or files? [unknown]: Which models accept images, audio or files? Unknown. Exact contracted/activated models, automatic cross-model routing and a complete image/audio/file matrix were not established. Catalogue dates include future release identifiers; dated additions and tenant eligibility take precedence over a release number.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:models, unique-ai:unique-ai:agent, unique-ai:unique-ai:conduct, unique-ai:unique-ai:release42.

MOD-02 Model freshness

Judgment rule: Measure the days from a provider's general release to availability in the product for the last three major releases from OpenAI, Anthropic and Google (and Mistral where offered). Use dated changelogs or observation. Record deprecation notice for retired models.

Unique AI

Agent cross-checked Support: partial

Unique publishes dated model additions and retirement/migration notes; a complete provider-release lag series remains unknown.

Limitations

A provider-verified last-three-major-release lag series for OpenAI, Anthropic and Google was not assembled; availability can depend on each contract. Admin version-pinning guarantees are not documented.

release_lag_days What was the lag for each of the last three major releases, per provider? [unknown]: What was the lag for each of the last three major releases, per provider? Unknown. A provider-verified last-three-major-release lag series for OpenAI, Anthropic and Google was not assembled; availability can depend on each contract. Admin version-pinning guarantees are not documented.
changelog Is there a dated public model changelog? [supported]: Availability overview has a dated Recent changes table through 13 October 2026, with provider retirement sources. Sources: unique-ai:unique-ai:models.
deprecation_notice How much notice is given before a model is removed? [partial]: Dated 14 October release announces migrations ahead of o-series provider retirement on 19 November 2026. No guaranteed Unique contractual notice period is established. Sources: unique-ai:unique-ai:release42.
version_pinning Can an admin keep an older model version? [unknown]: Can an admin keep an older model version? Unknown. A provider-verified last-three-major-release lag series for OpenAI, Anthropic and Google was not assembled; availability can depend on each contract. Admin version-pinning guarantees are not documented.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:models, unique-ai:unique-ai:release42.

MOD-03 Bring-your-own keys and provider contracts

Judgment rule: Record customer-held API keys, customer cloud accounts (Azure OpenAI, Bedrock, Vertex), and aggregator routes (OpenRouter, LiteLLM) as separate routes. Record what still works without a vendor-provided model, whether prices change with BYO, and documented restrictions. API-key support does not imply support for the customer's own Azure contract.

Unique AI

Agent cross-checked Support: partial

Operators can configure customer model endpoints and credentials through GraphQL.

Limitations

BYO licence discounts, provider-by-provider contract onboarding and a full tool/RAG/embedding parity matrix are not public.

own_api_keys Can the customer use its own provider API keys? [supported]: ExternalLanguageModel accepts provider API credentials stored encrypted as immutable write-only values; OAuth2 is also documented. Sources: unique-ai:unique-ai:customllm.
own_cloud_contract Can it use its own Azure OpenAI, Bedrock or Vertex deployment? [partial]: Custom Azure OpenAI deployment examples specify customer endpoint, deployment, API version and token. Compliance docs mention Bedrock; exact supported Bedrock/Vertex contract routes need deployment scoping. Sources: unique-ai:unique-ai:customllm, unique-ai:unique-ai:compliance.
custom_endpoint Can an admin add any OpenAI-compatible endpoint? [partial]: Documents self/cloud-hosted LXM endpoints and an Azure-compatible connection object. Universal support for any OpenAI-compatible endpoint is not guaranteed. Sources: unique-ai:unique-ai:customllm.
byo_pricing Does BYO change the licence price? [unknown]: Does BYO change the licence price? Unknown. BYO licence discounts, provider-by-provider contract onboarding and a full tool/RAG/embedding parity matrix are not public.
byo_restrictions Which features stop working with BYO? [partial]: Azure Chat Code Execution requires compatible Azure Responses models; switching a custom model group name/deleting an in-use group can break configured assistants. Sources: unique-ai:unique-ai:customllm, unique-ai:unique-ai:code.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:customllm, unique-ai:unique-ai:compliance, unique-ai:unique-ai:code.

MOD-04 Local and self-hosted models

Judgment rule: Record support for models running in the customer's environment (vLLM, Ollama, TGI or similar) and which features work with them: chat, RAG, tools, embeddings. Distinguish a documented, supported route from a generic endpoint that happens to work.

Unique AI

Agent cross-checked Support: partial

The platform documents custom model and embedding connections. Tool parity depends on the selected route: Code Execution lists Azure and LiteLLM OpenAI compatibility, while arbitrary self-hosted model parity is unverified.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration; version=not specified; plan=Contracted Unique AI platform; custom-model/embedding connectors and optional Code Execution availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

Code Execution has a listed Responses-capability/model matrix, not universal support for every custom-hosted model. Azure sandbox controls are not extended to every LiteLLM route.
No customer runtime or feature-parity fixture was used; QA-only Phoeniqs is not presented as GA.

self_hosted_chat Can chat use a customer-hosted model? [supported]: Custom LXM and on-premises architecture allow customer-hosted model connections. Models, endpoint compatibility and enabled modules are selected for the contracted deployment. Sources: unique-ai:unique-ai:customllm, unique-ai:unique-ai:onpremsecurity.
self_hosted_embeddings Can retrieval use customer-hosted embeddings? [partial]: An instance maintainer can configure embedding endpoint, key and deployment; the operator overview allows self-hosted embeddings. Supported non-Azure runtimes and a particular customer configuration were not verified. Sources: unique-ai:unique-ai:embeddings, unique-ai:unique-ai:operators.
feature_parity Which features fail with self-hosted models? [partial]: Changing embeddings requires re-embedding existing files. The Code Execution compatibility section explicitly lists Azure Responses-capable IDs and LiteLLM OpenAI IDs, so no blanket Azure-only restriction is asserted. Its introduction/security section describes an Azure sandbox; the actual LiteLLM execution route and parity for arbitrary self-hosted models remain unverified. Sources: unique-ai:unique-ai:embeddings, unique-ai:unique-ai:review2-code.
vendor_hosted_open_models Does the vendor host open-weight models in EU/CH? [partial]: The model catalogue describes Swiss Phoeniqs open models in QA only, not the UAT01/Prod offer. Production entitlement to a vendor-hosted Swiss open model was not established. Sources: unique-ai:unique-ai:review2-models.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:customllm, unique-ai:unique-ai:onpremsecurity, unique-ai:unique-ai:embeddings, unique-ai:unique-ai:operators, unique-ai:unique-ai:review2-code, unique-ai:unique-ai:review2-models.

MOD-05 Model control and usage limits

Judgment rule: Record admin allow-lists per group, default models, restrictions on which models may see which data, and per-user rate or fair-use limits. Test one restriction with two users in different groups. Record limits as enforced stops, warnings or undocumented fair use.

Unique AI

Agent cross-checked Support: partial

Models and knowledge/tools are configured per access-controlled Space; optional beta daily spending caps block calls.

Availability: beta · Delivery: native
Scope: surfaces=Unique AI Chat and administration; version=not specified; plan=Contracted Unique AI platform and enabled modules. No public named tariff tiers were established; deployment and provider entitlements vary by MSA.; add_ons=none stated; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

No two-group restriction fixture was run. Exact contract token/message limits, platform-wide model allow-lists and provider data-routing policy remain unverified.

model_allowlist_by_group Can admins limit models per group? [partial]: Admins bind deployed/approved models to Spaces and grant people/groups space access. Direct per-group model lists across all chats are not documented. Sources: unique-ai:unique-ai:agent.
data_bound_models Can admins restrict which models receive connector or knowledge content? [partial]: A Space binds one model, selected KB sources and tools, and people/group access. A global connector-content model policy is not established. Sources: unique-ai:unique-ai:agent.
default_model Can admins set the default model? [supported]: Administrators select the model for each Space. Sources: unique-ai:unique-ai:agent.
user_limits What per-user message or token limits apply, and are they published? [partial]: Off-by-default beta cost management supports daily USD limits at company/group/user levels, reset at user-local midnight. Public fixed token/message quotas are not established. Sources: unique-ai:unique-ai:cost.
limit_behaviour What happens when a limit is reached? [supported]: With limits enabled, the next model call is blocked when daily spend reaches the cap. Most permissive group limit wins unless a user override exists. Sources: unique-ai:unique-ai:cost.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:cost.

Company knowledge

KNOW-01 Files in chat

Judgment rule: Record supported formats, size and count limits, scanned-PDF OCR, tables and spreadsheets, images inside documents, and whether a file is read in full or retrieved in chunks. Upload the shared fixture set (long German PDF, scanned contract, Excel with formulas, PowerPoint) and record failures.

Unique AI

Agent cross-checked Support: partial

Chat upload defaults are 512 MB/file and 30 files/batch, with source retrieval for large documents.

Limitations

Configured scanned-PDF OCR is documented: MDI supports image-based PDFs, Docling has partial support, and Base Unique Ingestion does not support them. MDI must be provisioned and enabled for the folder or chat Space.
Scanned-PDF accuracy, formula preservation, embedded-figure extraction and shared fixture outcomes were not measured. Older general ingestion limits differ from current production chat defaults.

formats Which formats are accepted? [supported]: Chat supports PDF, DOCX, XLSX, PPTX, CSV, HTML, MD and TXT; PNG/JPEG when enabled per Space. Sources: unique-ai:unique-ai:files.
limits What size and count limits apply per file and chat? [partial]: Production Chat defaults 512 MB/file, 30 files/upload. Code Execution has separate 100 MB and 10-file limits. Environment overrides and aggregate chat caps need tenant confirmation. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:code.
ocr Are scanned PDFs read? [partial]: The file-ingestion guide explicitly distinguishes methods: Base Unique Ingestion does not support image-based scanned PDFs, MDI supports them, and Docling has partial support. MDI must be provisioned and enabled for the folder or in the chat Space ingestion settings. Experimental image-content extraction and deprecated agentic ingestion are separate. No scanned-document fixture or accuracy measurement was run. Sources: unique-ai:unique-ai:file-ingestion-20261026.
spreadsheets Are Excel tables read with structure intact? [partial]: Chat guide says default extraction can fail beyond 5,000 rows x 100 columns. General format guide lists 1,000 x 40; formulas and structure retention were not tested. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:filetypes.
full_vs_chunked Is the whole file used or only retrieved parts, and does the user see which? [supported]: Checked uploads force retrieval each turn. Large files contribute selected chunks, not necessarily the full file, and users do not see full coverage. Experimental full-PDF mode is separate. Sources: unique-ai:unique-ai:files.

Checked 2026-10-26; agent cross-checked 2026-10-26; check again by 2026-12-25. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:code, unique-ai:unique-ai:filetypes, unique-ai:unique-ai:file-ingestion-20261026.

KNOW-02 Knowledge bases and shared collections

Judgment rule: Record curated document collections that persist across chats, who can create and share them, size limits, and how they attach to assistants. Distinguish personal, group and organisation-wide scope.

Unique AI

Agent cross-checked Support: partial

Knowledge folders persist and can be scoped to users/groups and bound to Spaces.

Limitations

Public total collection/document/token caps and organisation-wide sharing defaults are not specified; chat-only uploads are separately private and expiring.

persistent_collections Can users keep document collections across chats? [supported]: Ingested content lives in knowledge-base folders/scopes beyond a chat. Sources: unique-ai:unique-ai:kbpermissions.
sharing_scope Can collections be shared with users, groups or the whole organisation? [supported]: ScopeAccess grants users or groups read/write access. A flat folder permission does not automatically grant subfolder access. Sources: unique-ai:unique-ai:kbpermissions.
size_limits What are the collection limits? [unknown]: What are the collection limits? Unknown. Public total collection/document/token caps and organisation-wide sharing defaults are not specified; chat-only uploads are separately private and expiring.
assistant_binding Can collections be bound to an assistant? [supported]: Space Internal Search binds selected folders/files and further narrows user permissions. Sources: unique-ai:unique-ai:agent.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:kbpermissions, unique-ai:unique-ai:agent.

KNOW-03 Enterprise connectors

Judgment rule: List each connector separately with plan, availability and setup owner: SharePoint, OneDrive, Outlook, Teams, Google Drive, Gmail, Confluence, Jira, Notion, Box, Salesforce, ServiceNow, file shares. Record whether the vendor or a third party (or the customer via MCP) provides it, and whether it needs admin consent. A connector listed on a roadmap or partner page is `announced`. An on-prem connector or gateway proves access to local data, not a self-hosted platform (DATA-02).

Unique AI

Agent cross-checked Support: partial

Indexed SharePoint and Confluence are documented. Current architecture also names Google Drive/Salesforce, and release notes describe Salesforce/Yahoo Finance MCP connector dashboards; detailed entitlements and source permissions remain connector-specific.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique MCP Hub and Knowledge Base MCP; version=not specified; plan=Contracted Unique AI platform and enabled modules. No public named tariff tiers were established; deployment and provider entitlements vary by MSA.; add_ons=none stated; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

Architecture labels and release examples do not prove that every source is enabled or included in every contract.
Future items on the roadmap are not promoted to current connector support.
No source-connection or two-user permission fixture was run.

microsoft_365 Which M365 sources (SharePoint, OneDrive, Outlook, Teams) are connected? [partial]: SharePoint Online connector copies selected content with application consent and optional permission sync. Platform documentation also exposes Teams and Outlook MCP operator guides, but their exact contracted scope and OneDrive coverage were not verified. Sources: unique-ai:unique-ai:sharepoint.
google_workspace Which Google sources are connected? [partial]: The current Platform Overview names Google Drive as a Knowledge Source feeding the Knowledge Base. It does not establish Gmail/Calendar coverage, a deployment guide, source consent scopes or inclusion in a particular customer contract. This is a documented architecture listing, not an observed enabled connector. Sources: unique-ai:unique-ai:platform-overview-20261026.
atlassian Confluence and Jira? [partial]: Official Confluence connector supports Cloud and Data Center; Jira connector coverage was not established. Sources: unique-ai:unique-ai:confluence.
other_systems Which other business systems? [partial]: Release 2026.40 names Salesforce and Yahoo Finance as examples in GA MCP connector dashboards that admins can list, enable and sign in to. Platform Overview also names Salesforce as a Knowledge Source. Exact origin implementation, connector rights, credentials and commercial inclusion vary by configured connector and contract. Sources: unique-ai:unique-ai:release40-20261026, unique-ai:unique-ai:platform-overview-20261026.
on_prem_sources Can on-prem file shares or databases be connected? [partial]: Confluence Data Center sources are supported via REST. Generic file-share/database connector scope was not established. Sources: unique-ai:unique-ai:confluence.
connector_provider Who builds and maintains each connector? [supported]: Unique publishes Kubernetes operator guides for SharePoint and Confluence; client IT supplies source credentials, consent, scope and deployment configuration. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence.

Checked 2026-10-26; agent cross-checked 2026-10-26; check again by 2026-12-25. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence, unique-ai:unique-ai:platform-overview-20261026, unique-ai:unique-ai:release40-20261026.

KNOW-04 Permission-aware retrieval

Judgment rule: Using two test users, verify that a restricted document in the fixture tenant is never cited or summarised for the user without access, including after a permission change. Record whether permissions are enforced at query time by the source system, mirrored into an index, or not enforced. Record the propagation delay of a revocation.

Unique AI

Agent cross-checked Support: partial

SharePoint permission mirroring is optional; Confluence does not synchronise source permissions.

Limitations

No two-user restricted-document/revocation fixture was run. End-to-end revocation SLA and label enforcement are not established. Older folder-permissions documentation says SharePoint never mirrors permissions, conflicting with the newer optional-sync guide.

source_permissions_respected Are source permissions enforced per user? [partial]: SharePoint content_and_permissions mirrors source rights, while content_only uses Unique scope rules. Confluence explicitly has no source-permission sync, so Unique folder access needs configuration. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence.
enforcement_point Does the source system, a mirrored index or nothing enforce them? [supported]: Indexed content is filtered through Unique user/group ScopeAccess. This is not source-system live query-time enforcement. Sources: unique-ai:unique-ai:kbpermissions, unique-ai:unique-ai:sharepoint.
revocation_delay How quickly does a removed permission take effect? [partial]: SharePoint source rights refresh periodically, default scans 15 minutes. Internal gatekeeper changes usually propagate under one second but have no hard upper bound; degraded/missed broadcasts can leave old rules active. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:permissioncache.
admin_scope_controls Can admins exclude sites, folders or labels? [supported]: Admins allow SharePoint sites/libraries and require a sync flag, Confluence labels, and Space KB restrictions. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence, unique-ai:unique-ai:agent.
sensitivity_labels Are Microsoft Purview or similar labels respected? [partial]: July SharePoint connector explicitly excludes sensitivity/classification-level detection. Separately, release2026.42 documents experimental Purview-protected-file decryption using uploader rights, gated by MIP_DECRYPTION_ENABLED and Entra/operator changes. This is not proof of general label-policy enforcement. Sources: unique-ai:unique-ai:release42, unique-ai:unique-ai:sharepoint.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence, unique-ai:unique-ai:kbpermissions, unique-ai:unique-ai:permissioncache, unique-ai:unique-ai:agent, unique-ai:unique-ai:release42.

KNOW-05 Freshness and sync model

Judgment rule: Record live (federated) search versus a copied index, sync interval, deletion propagation, and where the index is stored. Edit and delete one fixture document and record when answers change. A copied index adds storage location to DATA-01.

Unique AI

Agent cross-checked Support: partial

SharePoint and Confluence copy content to Unique indexes on configurable scans, default every 15 minutes.

Limitations

No source-edit/delete fixture was run. Actual sync duration, strict deletion SLA and index/backup region per contract remain unverified.

live_or_indexed Is content searched live or copied into an index? [supported]: Connectors ingest documents/page attachments into the Unique knowledge base for RAG. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence.
sync_interval How often does an index refresh? [supported]: Default 15-minute scan for both connectors; schedules are configurable. Real-time sync is explicitly not supported. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence.
deletion_propagation How quickly does a deleted document disappear? [partial]: Diff processing detects deleted source files/pages and removes them on a sync run. A strict completion bound is not provided. Sources: unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:confluence.
index_location Where is indexed content stored? [partial]: Hosted documents use Azure storage; embeddings Qdrant and search vectors on managed disks in the selected deployment. Exact per-client index region needs the MSA/architecture. Sources: unique-ai:unique-ai:retention, unique-ai:unique-ai:keys.

KNOW-06 Grounding and retrieval quality

Judgment rule: Run the shared question set over the fixture corpus: factual lookups, cross-document questions, German document with an English question, a question with no answer in the corpus. Record citation presence, link accuracy, correct answers and refusals to invent. Disclose model, settings, attempt count and failures. Single runs support a case study, not a ranking.

Unique AI

Agent cross-checked Support: partial

Unique documents source-linked answers and reference limits, without a shared-corpus accuracy result.

Limitations

No shared question set was run. Citation correctness, cross-language recall, no-answer behaviour and observed accuracy remain unmeasured. Vendor evaluation tools are not our benchmark results.

citations Are answers linked to specific sources? [supported]: Chat retrieval cites sources; Conduct links source facts, with PDF/Word page-level references and other formats at document level. Generated files are not citation-linked or hallucination-checked. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:conduct.
citation_accuracy Do the cited passages support the answer? [unknown]: Do the cited passages support the answer? Unknown. No shared question set was run. Citation correctness, cross-language recall, no-answer behaviour and observed accuracy remain unmeasured. Vendor evaluation tools are not our benchmark results.
cross_language Does a question in one language find documents in another? [unknown]: Does a question in one language find documents in another? Unknown. No shared question set was run. Citation correctness, cross-language recall, no-answer behaviour and observed accuracy remain unmeasured. Vendor evaluation tools are not our benchmark results.
no_answer_behaviour Does it say when the corpus has no answer? [unknown]: Does it say when the corpus has no answer? Unknown. No shared question set was run. Citation correctness, cross-language recall, no-answer behaviour and observed accuracy remain unmeasured. Vendor evaluation tools are not our benchmark results.
observed_accuracy How many of the question set were answered correctly? [unknown]: How many of the question set were answered correctly? Unknown. No shared question set was run. Citation correctness, cross-language recall, no-answer behaviour and observed accuracy remain unmeasured. Vendor evaluation tools are not our benchmark results.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:conduct.

KNOW-07 Memory and personal context

Judgment rule: Record personal custom instructions, cross-chat memory, project or workspace context, and user controls to view, edit, delete or disable memory. Record whether admins can disable memory organisation-wide.

Unique AI

Agent cross-checked Support: partial

Context Memory provides a private cross-chat user profile with edit and opt-out controls.

Limitations

No authenticated memory fixture or permanent deletion test was run. The guide explicitly contradicts the settings deletion promise.

custom_instructions Can users set persistent instructions? [partial]: Admins configure Space instructions; memory learns preferences. A separate persistent user-authored global instruction field was not established. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:memory.
cross_chat_memory Does the product remember facts across chats? [supported]: One per-user profile is loaded across enabled spaces/chats and updated after answers; stored in a private unindexed KB folder. Sources: unique-ai:unique-ai:memory.
user_memory_controls Can users view, edit and delete memories? [partial]: Users view/edit the profile and disable updates/loading. Docs explicitly say disabling does not delete the profile despite contrary settings text; a delete control is not established. Sources: unique-ai:unique-ai:memory.
admin_memory_controls Can admins turn memory off? [supported]: Release2026.42 makes Context Memory GA, with per-company enable/disable feature flag and per-Space activation. Users can opt out; Web Search combinations remain restricted by default. Sources: unique-ai:unique-ai:release42, unique-ai:unique-ai:memory.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:memory, unique-ai:unique-ai:release42.

Assistants and agents

AGT-01 Custom assistants

Judgment rule: Record creation of assistants with instructions, model, knowledge and tools; who can create; sharing scope; versioning; and an internal catalogue or directory. Build one HR policy assistant on the fixture and share it with one group. Prompt templates and a prompt library are recorded here.

Unique AI

Agent cross-checked Support: partial

Admins configure Spaces with instructions, models, knowledge, tools, prompts and group access.

Limitations

Assistant versioning, organisation-wide directory discovery and unrestricted end-user creation were not established.

create_assistants Can users create assistants with instructions, knowledge and tools? [supported]: Space setup includes model, instructions, knowledge sources and tools. Sources: unique-ai:unique-ai:agent.
creator_permissions Can admins limit who creates them? [supported]: Space creation/management requires delegated role permissions; Can manage grants space configuration access. Sources: unique-ai:unique-ai:roles, unique-ai:unique-ai:agent.
sharing_scope Can assistants be shared with users, groups or everyone? [supported]: Share through Access with people/groups as Can chat or Can manage. Sources: unique-ai:unique-ai:agent.
catalogue Is there an internal assistant directory? [partial]: Space Management lists configured workspaces; a dedicated employee assistant-discovery directory is not described. Sources: unique-ai:unique-ai:agent.
prompt_library Are shared prompt templates available? [supported]: Admins publish Suggested Prompts with a Space; users also have personal saved prompts. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:chat.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:roles, unique-ai:unique-ai:chat.

AGT-02 Tools, actions and MCP

Judgment rule: Record built-in tools, custom actions (OpenAPI), MCP client support (local and remote servers), authentication model per tool (shared service account or per-user OAuth), and admin approval of tools. Verify one external-tool call and its access boundary.

Unique AI

Agent cross-checked Support: partial

Built-in tools and beta MCP connectors are governed at organisation and Space levels.

Availability: beta · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique AI Conduct, Unique MCP Hub and Knowledge Base MCP; version=not specified; plan=Contracted platform with MCP beta feature flag and configured connectors; add_ons=MCP entitlement and enablement per contract; deployment_mode=not specified; region=Tenant plus downstream connector/provider regions; os=not specified.

Limitations

Arbitrary OpenAPI action import, local stdio MCP support, each connector auth mode and universal side-effect approval are not established. Hub planned DLP/elicitation pipeline is not treated as released.

builtin_tools Which tools are built in? [supported]: Internal search, web search, Deep Research and Code Execution are documented; Deep Research is experimental and code has model restrictions. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:code, unique-ai:unique-ai:deepresearch.
custom_actions Can admins add API actions? [unknown]: Can admins add API actions? Unknown. Arbitrary OpenAPI action import, local stdio MCP support, each connector auth mode and universal side-effect approval are not established. Hub planned DLP/elicitation pipeline is not treated as released.
mcp_client Can it use remote MCP servers, and with what admin control? [supported]: MCP Tools for Spaces is beta, off until enabled for the organisation; connector admins and space managers can selectively expose connectors/tools. Sources: unique-ai:unique-ai:mcp.
tool_auth Do tools act as the user or as a shared account? [supported]: MCP Hub maps user hub tokens to downstream per-user OAuth tokens; secrets are injected server-side. Sources: unique-ai:unique-ai:hub.
tool_approval Can admins approve or block tools, and can users confirm actions before they run? [partial]: Admins restrict origin/connector/tool exposure. Conduct can pause for approve/deny of configured sensitive actions; it is not a universal confirmation guarantee. Sources: unique-ai:unique-ai:hub, unique-ai:unique-ai:conduct.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:code, unique-ai:unique-ai:deepresearch, unique-ai:unique-ai:mcp, unique-ai:unique-ai:hub, unique-ai:unique-ai:conduct.

AGT-03 Web search and deep research

Judgment rule: Record web search on/off control per organisation and group, search provider, citation of web sources, deep or multi-step research, and whether queries leave the residency boundary. Admins of regulated companies often need web search off.

Unique AI

Agent cross-checked Support: partial

Web search has provider/strategy controls; separate Deep Research is experimental.

Limitations

No search/report fixture was run. Exact query/provider regions, contract enablement and current Deep Research production readiness remain unverified.

web_search Is web search available? [supported]: Real-time web search/read is a Space tool, with stable V1/V2 modes and experimental V3. Sources: unique-ai:unique-ai:websearch.
admin_toggle Can admins disable it per organisation or group? [supported]: Admins enable/disable tools per Space shared with groups. Deep Research tools are configured separately; disabling ordinary Web Search is not a universal external-search block. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:deepresearch.
search_provider Which search provider handles queries, and where? [partial]: Configured engines can include Google, Bing, Brave, Jina, Tavily, Firecrawl, Vertex or custom API; query context can leave the tenant. Exact provider region must be set per contract. Sources: unique-ai:unique-ai:websearch.
deep_research Is multi-step research with a report available? [supported]: Experimental tool creates multi-source research and reports. Unique engine can use internal/web sources; OpenAI-native engine requires direct OpenAI-hosted models and lacks internal sources/hosting-location control. Sources: unique-ai:unique-ai:deepresearch.
web_citations Are web sources cited? [supported]: Web answers include source references; Deep Research documents citation-linked reports. Sources: unique-ai:unique-ai:websearch, unique-ai:unique-ai:deepresearch.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:websearch, unique-ai:unique-ai:agent, unique-ai:unique-ai:deepresearch.

AGT-04 Workflows and automation

Judgment rule: Record no-code or low-code workflows, scheduled prompts, triggers from email, forms or business systems, and multi-step chains. Name where they run and who can build them. A shared prompt template is not a workflow.

Unique AI

Agent cross-checked Support: partial

Recurring Scheduled Tasks run with user permissions; current release notes label the feature experimental.

Limitations

No general no-code workflow/chain builder or complete event-source trigger catalogue was established. Automated runs may stop for manual approval/questions and are subject to normal usage limits.

workflow_builder Is there a workflow or chain builder? [unknown]: Is there a workflow or chain builder? Unknown. No general no-code workflow/chain builder or complete event-source trigger catalogue was established. Automated runs may stop for manual approval/questions and are subject to normal usage limits.
scheduled_tasks Can prompts or assistants run on a schedule? [partial]: User guide labels recurring Chat/Conduct tasks beta and off by default; release2026.42 instead labels Scheduled Tasks experimental, off by default, enabled on request, and describes scheduling/eligibility fixes. Sources: unique-ai:unique-ai:release42, unique-ai:unique-ai:scheduled.
event_triggers Can external events start a workflow? [partial]: SDK documents webhook-driven applications, which require developer-hosted modules; a packaged no-code external-event builder was not established. Sources: unique-ai:unique-ai:sdk.
builder_audience Is building aimed at business users, IT or developers? [partial]: Business users author recurring tasks in the UI or Conduct; developers host webhook-driven SDK modules. Availability and scheduling behaviour depend on deployed release. Sources: unique-ai:unique-ai:scheduled, unique-ai:unique-ai:sdk.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:scheduled, unique-ai:unique-ai:sdk, unique-ai:unique-ai:conduct, unique-ai:unique-ai:release42.

AGT-05 Agentic task execution

Judgment rule: Record an agent mode that plans and carries out multi-step tasks with tools (browser, files, connectors), with visible steps, user confirmation of side effects and stop control. Run one fixture task (gather three documents, draft a summary, propose an email). Record interventions and what ran. A final claim of success is not proof.

Unique AI

Agent cross-checked Support: partial

Contract-enabled Conduct plans and executes multi-step tasks with visible tools and stop controls. Sensitive actions can ask for approval; separately enabled Experimental Scheduled Tasks offer automatic approval.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Conduct; version=Conduct 2026.40+; scheduling status/flags per 2026.42; plan=Contracted Unique AI platform; Conduct; Scheduled Tasks separately enabled per contract and feature flag availability and deployment/provider configuration depend on the customer agreement; add_ons=Conduct; optional Scheduled Tasks feature flag; deployment_mode=not specified; region=Configured tenant/models/tools; scheduled actions run with the owner permissions and configured approval policy; os=not specified.

Limitations

No agent fixture was run. Interactive sensitive-action approval and separately enabled unattended automatic approval are different execution modes.
Actual tools, entitlement, approval policy and deployed release depend on the customer contract/configuration.

agent_mode Is there an agent mode for multi-step tasks? [supported]: Plans and executes Python/Bash, files, internal search and configured MCP tools in an autonomous workspace. Sources: unique-ai:unique-ai:conduct.
visible_steps Are steps and tool calls shown? [supported]: Shows checklist, progress, code/tool inputs and outputs, and waiting/working indicators. Sources: unique-ai:unique-ai:conduct.
side_effect_confirmation Must the user confirm actions with side effects? [partial]: Conduct shows approve/deny for actions configured as sensitive. The separate Scheduled Tasks guide offers Automatic or Manual approval for Chat/Conduct unattended runs. Release 2026.42 labels scheduling Experimental and feature-flagged. Mandatory human approval of every side effect is therefore not established for all configured runs. Sources: unique-ai:unique-ai:conduct, unique-ai:unique-ai:review2-scheduled, unique-ai:unique-ai:review2-release42.
stop_control Can the user stop a running task? [supported]: Stop ends the turn within seconds; completed steps/files remain and interrupted work can be resumed. Sources: unique-ai:unique-ai:conduct.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:conduct, unique-ai:unique-ai:review2-scheduled, unique-ai:unique-ai:review2-release42.

AGT-06 Programmable access

Judgment rule: Record a public API for chat, assistants and knowledge, SDKs, API authentication, rate limits and whether API use counts against the same governance (audit, budgets, model restrictions). Being an MCP client is different from exposing the product over MCP.

Unique AI

Agent cross-checked Support: partial

Public API/Python SDK and Knowledge Base MCP are documented.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique API and Python SDK, Unique MCP Hub and Knowledge Base MCP; version=not specified; plan=Contracted APIs and KB MCP deployment; model tracking requires enablement; add_ons=KB MCP and usage-tracking enablement per contract; deployment_mode=not specified; region=Configured deployment and external client region; os=not specified.

Limitations

No API/SDK fixture was run; complete endpoint catalogue, rate limits, budget enforcement and all-services audit parity remain unverified.

chat_api Is there an API for chat and assistants? [partial]: SDK exposes the Unique public API, personal/app keys and webhook signatures. Exact endpoint entitlements/rate limits require the contracted deployment. Sources: unique-ai:unique-ai:sdk.
knowledge_api Can documents be loaded through an API? [partial]: Connector/operator docs describe ingestion and knowledge APIs; public SDK supports platform integrations. Sources: unique-ai:unique-ai:sdk, unique-ai:unique-ai:sharepoint.
api_governance Do audit, budgets and model rules apply to API use? [partial]: Feature-flagged model-use tracking covers public API/SDK/agentic and Azure SDK calls with company/user/app metadata. Complete budget/model-rule enforcement across all internal services is not established. Sources: unique-ai:unique-ai:api-usage.
mcp_server Does the product expose itself to other agents over MCP? [supported]: Knowledge Base MCP lets external assistants search/read the Unique KB with the user permissions and admin scope restrictions. The documentation-only MCP endpoint is distinct. Sources: unique-ai:unique-ai:mcpkb.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:sdk, unique-ai:unique-ai:sharepoint, unique-ai:unique-ai:api-usage, unique-ai:unique-ai:mcpkb.

Daily work

WORK-01 Clients and access points

Judgment rule: List web, desktop (Windows, macOS), iOS, Android, mobile browser and browser extension separately. For each, record login with company SSO, feature gaps against web, and MDM/Intune deployability. An installable app is different from a mobile website.

Unique AI

Agent cross-checked Support: partial

A web workspace is documented; native-client and managed-deployment coverage is unverified.

Limitations

The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.

web Is a full web client available? [supported]: Public documentation describes browser chat, Space Management and administration. Sources: unique-ai:unique-ai:chat, unique-ai:unique-ai:agent.
desktop_apps Windows and macOS apps? [unknown]: Windows and macOS apps? Unknown. The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.
mobile_apps iOS and Android apps with SSO? [unknown]: iOS and Android apps with SSO? Unknown. The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.
browser_extension Is there a browser extension or sidebar? [unknown]: Is there a browser extension or sidebar? Unknown. The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.
managed_deployment Can IT deploy and configure clients through MDM? [unknown]: Can IT deploy and configure clients through MDM? Unknown. The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.
client_feature_gaps Which features are missing on mobile? [unknown]: Which features are missing on mobile? Unknown. The public material references an app and mobile voice/browser usage without a current Windows/macOS/iOS/Android SSO/MDM matrix. Browser extension/sidebar entitlement and mobile parity remain unknown.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:chat, unique-ai:unique-ai:agent.

WORK-02 Office suite and Teams integration

Judgment rule: Record presence inside Word, Excel, PowerPoint, Outlook and Teams (and Google Docs, Gmail), with what the assistant can do in each: read the open document, write into it, act on mail. Record extra licences and admin consent needed. A connector that searches Office files is KNOW-03, not WORK-02.

Unique AI

No public information found Support: not-published

We found no public information on this. A missing finding is not evidence that the product lacks the capability.

WORK-03 Translation and writing

Judgment rule: Record document translation preserving layout, the translation engine (LLM or dedicated such as DeepL), glossaries, formality control, supported languages including Swiss German conventions (no ß), and rewriting/proofreading tools. Translate the shared German contract into French and Italian and record layout and glossary results.

Unique AI

Agent cross-checked Support: partial

A beta Translator Space translates Office files and text with configurable glossaries.

Limitations

No translated-contract fixture was run. Exact engine/model, formality/de-CH controls, supported glossary language pairs and layout fidelity need verification.

document_translation Can whole documents be translated with formatting kept? [partial]: DOCX, XLSX and PPTX document translation is documented, with possible layout disruption. PDF input is explicitly excluded in this Translator mode. Sources: unique-ai:unique-ai:translation.
translation_engine Which engine translates? [partial]: Translator V2 is a standalone AI module configured with a model; no dedicated DeepL engine was established. Sources: unique-ai:unique-ai:translation.
glossaries Are company glossaries supported? [supported]: Admins can configure glossary pre/postprocessing. Sources: unique-ai:unique-ai:translation.
formality_and_variants Can formality and regional variants (de-CH) be set? [partial]: Postprocessing can change regional spelling, exemplified by US-to-UK English. Formality or de-CH-specific support was not verified. Sources: unique-ai:unique-ai:translation.
writing_tools Are dedicated rewrite and proofreading tools available? [partial]: Chat generates and iterates on text; no distinct proofreading/rewriting tool scope was established. Sources: unique-ai:unique-ai:chat.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:translation, unique-ai:unique-ai:chat.

WORK-04 Image generation and editing

Judgment rule: Record available image models, editing (inpainting, reference images), admin control and per-image cost visibility, and licence or usage terms for generated images.

Unique AI

No public information found Support: not-published

We found no public information on this. A missing finding is not evidence that the product lacks the capability.

WORK-05 Voice and meetings

Judgment rule: Record dictation, read-aloud, live voice conversation, meeting transcription and summarisation, supported languages including Swiss German speech, and where audio is processed.

Unique AI

Agent cross-checked Support: partial

Voice dictation is documented with Azure Speech prerequisites; recording use is separately described in the DPA.

Limitations

Read-aloud and full voice conversation, recording-specific plans/limits and verified Swiss German recognition are not established. Voice guide warns about dialects; audio route may traverse public internet/Azure backbone.

dictation Can users dictate prompts? [supported]: Microphone transcribes into the prompt field, with audio discarded after live transcription. Requires configured Azure AI Speech Services; EN/DE/FR/IT are listed. Sources: unique-ai:unique-ai:voice, unique-ai:unique-ai:voice-user.
read_aloud Can answers be read aloud? [unknown]: Can answers be read aloud? Unknown. Read-aloud and full voice conversation, recording-specific plans/limits and verified Swiss German recognition are not established. Voice guide warns about dialects; audio route may traverse public internet/Azure backbone.
voice_conversation Is a live voice mode available? [unknown]: Is a live voice mode available? Unknown. Read-aloud and full voice conversation, recording-specific plans/limits and verified Swiss German recognition are not established. Voice guide warns about dialects; audio route may traverse public internet/Azure backbone.
meeting_summaries Are meetings transcribed and summarised? [partial]: DPA recording use case covers audio/transcripts, optional video and GPT summaries/follow-up emails. This is a contracted recording use case, not proof every Chat plan includes meeting capture. Sources: unique-ai:unique-ai:dpa.
swiss_german_speech Is Swiss German speech recognised? [unknown]: Is Swiss German speech recognised? Unknown. Read-aloud and full voice conversation, recording-specific plans/limits and verified Swiss German recognition are not established. Voice guide warns about dialects; audio route may traverse public internet/Azure backbone.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:voice, unique-ai:unique-ai:voice-user, unique-ai:unique-ai:dpa.

WORK-06 Data analysis and document output

Judgment rule: Record code execution for data analysis, charts, and generation of downloadable Word, Excel, PowerPoint and PDF files. Analyse the fixture Excel and request a PowerPoint summary. Record where code runs and whether files are editable.

Unique AI

Agent cross-checked Support: partial

Enabled Chat Code Execution and Conduct can analyse data and produce downloadable files. Selected KB documents can be preloaded into the code container; ordinary unrestricted KB retrieval is separate.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique AI Conduct; version=not specified; plan=Contracted Unique AI platform; Chat Code Execution or Conduct; model compatibility and flags configured separately availability and deployment/provider configuration depend on the customer agreement; add_ons=Code Execution / Conduct enablement per contract; deployment_mode=not specified; region=Documented Azure code container for the Azure route; LiteLLM execution region unverified; Conduct sandbox per contracted deployment; os=not specified.

Limitations

Selected admin-preloaded KB files are supported; this is not arbitrary sandbox retrieval of the whole Knowledge Base.
The page lists Azure and LiteLLM-compatible model IDs but describes Azure-specific container controls. Actual provider execution routing and entitlement require deployment configuration.
No generated-file quality/editability fixture was run.

data_analysis Can it analyse spreadsheets with executed code? [partial]: Chat Code Execution runs Python for CSV/XLSX analysis. The general limitation says KB files are not directly accessible, but admin configuration additionalUploadedDocuments preloads selected KB content_ids, for example templates, into the container. The documented Azure route has no outbound internet/runtime installs and a ten-file/100 MB per-file limit. The compatibility table also lists LiteLLM OpenAI models; that does not establish the same execution-region controls for every route. Conduct separately runs Python/Bash. Sources: unique-ai:unique-ai:review2-code, unique-ai:unique-ai:conduct.
charts Can it produce charts? [supported]: The Code Execution guide documents matplotlib PNG/SVG and inline charts. Its described Azure container lacks kaleido, limiting Plotly static exports; no statement about every alternative runtime is inferred. Sources: unique-ai:unique-ai:review2-code.
office_file_output Can it create Word, Excel or PowerPoint files? [partial]: The guide documents DOCX/XLSX generation and PPTX from supplied templates, with variable presentation quality. KB templates can be supplied by the admin preload setting. Generated files remain outside Conduct source-citation/hallucination checks. Sources: unique-ai:unique-ai:review2-code, unique-ai:unique-ai:conduct.
canvas_editing Is there a document or canvas editor for iterating on output? [unknown]: No dedicated canvas editor or editable spreadsheet-to-PowerPoint fixture result was established. Downloadable file generation does not establish in-product canvas editing.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-code, unique-ai:unique-ai:conduct.

WORK-07 Sharing and team spaces

Judgment rule: Record chat sharing (link, user, group), shared projects or team spaces, channels or group chat with AI, and commenting. Test sharing a chat with a colleague in another group and record what the recipient sees, including knowledge they lack access to.

Unique AI

Agent cross-checked Support: partial

Shared Chats is GA in release 2026.42, with company settings and a per-Space toggle. Simultaneous group conversation and recipient source-permission handling remain unverified.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration; version=2026.42 or later; plan=Contracted Unique AI platform; Shared Chats and Spaces; sharing settings and deployed release availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

A shared conversation is not evidence of simultaneous group participation or recipient access to every source.
No cross-group shared-chat permission fixture was run.

share_chats Can chats be shared inside the organisation? [supported]: Release 2026.42 makes whole-chat sharing GA and on by default for companies, while each Space still needs its Shared chat toggle. Explicit existing company preferences are preserved. Owners can search recipients and copy share links. Sources: unique-ai:unique-ai:review2-release42.
team_spaces Are shared projects or workspaces available? [partial]: Admins publish shared Spaces with people/group Can chat and Can manage roles. These share agent configuration/knowledge, not necessarily a conversation. Sources: unique-ai:unique-ai:agent.
group_chat Can several people chat together with the AI? [unknown]: Whole-chat sharing is documented. A simultaneous multi-person AI conversation, collaborative commenting or live co-editing mode was not established or tested.
share_permission_boundary Does sharing respect the recipient's access to sources? [unknown]: Whole-chat sharing is documented, but whether a recipient can open every cited source or inherits source permissions was not established in the retrieved release/Space documentation. No cross-group fixture was run.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:review2-release42.

WORK-08 Interface languages and accessibility

Judgment rule: Record UI languages, focusing on German, French, Italian and English, with translation quality observed (complete, partial, machine-translated). Record response language handling and WCAG conformance statements.

Unique AI

Agent cross-checked Support: partial

Current release notes document German, French and Swiss German UI language behaviour; completeness and accessibility remain unverified.

Limitations

Release notes describe locale fixes, not a full UI audit or conformance statement. No authenticated UI was reviewed.

ui_languages Which UI languages are offered? [partial]: Release2026.42 describes German/French label translations and selected-language login/notices/errors/side-panels, including Swiss German spellings. A complete language matrix was not provided. Sources: unique-ai:unique-ai:release42.
dach_language_quality Are German, French and Italian UIs complete and natural? [unknown]: Are German, French and Italian UIs complete and natural? Unknown. DE/FR/IT/EN voice and text-language capabilities are not a UI-localisation matrix. No authenticated user/admin language-quality pass or current WCAG conformance statement was obtained.
accessibility_statement Is there a WCAG conformance statement? [unknown]: Is there a WCAG conformance statement? Unknown. DE/FR/IT/EN voice and text-language capabilities are not a UI-localisation matrix. No authenticated user/admin language-quality pass or current WCAG conformance statement was obtained.
admin_ui_languages Is the admin interface available in German? [unknown]: Is the admin interface available in German? Unknown. DE/FR/IT/EN voice and text-language capabilities are not a UI-localisation matrix. No authenticated user/admin language-quality pass or current WCAG conformance statement was obtained.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:release42.

Administration and governance

ADM-01 Single sign-on and authentication

Judgment rule: Record SAML, OIDC, Microsoft Entra ID, Google, LDAP, MFA enforcement, conditional-access compatibility, domain verification and plan restrictions. Test one SSO login on a trial where possible.

Unique AI

Agent cross-checked Support: partial

Zitadel supports Entra OIDC, generic OIDC and SAML across documented deployment types.

Limitations

Entra conditional-access compatibility, verified domain capture and tariff-specific SSO/MFA entitlements were not tested/documented.

sso_protocols Which SSO protocols and identity providers work? [supported]: Setup guides explicitly cover Entra ID OIDC, generic OIDC and SAML, including IdP metadata for closed networks. Sources: unique-ai:unique-ai:sso.
sso_plan Which plan includes SSO? [partial]: Compliance matrix lists SSO for multi-tenant, single-tenant, client-hosted and on-premise modes. Numeric tariff prerequisites are not public. Sources: unique-ai:unique-ai:sso, unique-ai:unique-ai:compliance.
mfa Can MFA be enforced? [partial]: Hosted compliance matrix lists enforced 2FA; SSO setup delegates authentication to the external IdP. Exact customer MFA policy controls need verification. Sources: unique-ai:unique-ai:sso, unique-ai:unique-ai:compliance.
conditional_access Do Entra conditional-access policies apply? [unknown]: Do Entra conditional-access policies apply? Unknown. Entra conditional-access compatibility, verified domain capture and tariff-specific SSO/MFA entitlements were not tested/documented.
domain_capture Can the company claim its email domain? [unknown]: Can the company claim its email domain? Unknown. Entra conditional-access compatibility, verified domain capture and tariff-specific SSO/MFA entitlements were not tested/documented.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:sso, unique-ai:unique-ai:compliance.

ADM-02 User provisioning and lifecycle

Judgment rule: Record SCIM, just-in-time provisioning, group sync from the identity provider, deprovisioning effects (chats, assistants, knowledge ownership transfer) and licence assignment. Deprovision a test user and record what happens to their content.

Unique AI

Agent cross-checked Support: partial

SCIM, group sync and SAML automatic user creation are documented.

Limitations

Leaver-content ownership transfer, deletion effects and licence assignment/reclamation are not specified; no deprovisioning fixture was run.

scim Is SCIM provisioning supported? [supported]: Unique supplies SCIM URL/key for IdP synchronisation, with an Entra guide. Sources: unique-ai:unique-ai:scim.
group_sync Are identity-provider groups synced? [supported]: SCIM syncs users/groups; SSO group claims can also sync on login, requiring optional Graph group permission/admin consent. Sources: unique-ai:unique-ai:scim, unique-ai:unique-ai:sso.
jit_provisioning Are users created at first login? [supported]: SAML setup explicitly enables automatic creation/update and account creation. Sources: unique-ai:unique-ai:sso.
deprovisioning What happens to a removed user's data and shared assets? [unknown]: What happens to a removed user's data and shared assets? Unknown. Leaver-content ownership transfer, deletion effects and licence assignment/reclamation are not specified; no deprovisioning fixture was run.
seat_assignment Can admins assign and reclaim seats? [unknown]: Can admins assign and reclaim seats? Unknown. Leaver-content ownership transfer, deletion effects and licence assignment/reclamation are not specified; no deprovisioning fixture was run.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:scim, unique-ai:unique-ai:sso.

ADM-03 Roles, groups and feature permissions

Judgment rule: Record admin roles, delegated admins, groups, per-feature permissions (web search, image generation, connectors, file upload, agents), separation of subsidiaries or departments, and enforcement. Test one disabled feature for one group.

Unique AI

Agent cross-checked Support: partial

Four role/access layers separate identity, feature, Space and knowledge controls.

Limitations

No two-user disabled-feature test was run. Full cross-Space per-group controls and legal-entity separation require contracted setup.

admin_roles Which admin roles exist, and can admin rights be delegated? [supported]: IAM owner/viewer/org-manager plus Unique application roles delegate administration; Space Can manage grants local configuration. Sources: unique-ai:unique-ai:roles, unique-ai:unique-ai:agent.
groups Can permissions be set per group? [supported]: Users/groups receive Space and ScopeAccess permissions. Sources: unique-ai:unique-ai:roles.
feature_toggles Which features can be switched on or off per group? [partial]: Tools and upload settings are configured per Space, then Spaces shared to groups. A global group-feature matrix is not established. Sources: unique-ai:unique-ai:agent.
multi_entity Can subsidiaries or departments be separated within one contract? [partial]: Compliance docs describe workspace isolation for departments/client engagements. Multi-subsidiary contractual/billing boundaries are not specified. Sources: unique-ai:unique-ai:compliance.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:roles, unique-ai:unique-ai:agent, unique-ai:unique-ai:compliance.

ADM-04 Usage and adoption analytics

Judgment rule: Record dashboards for active users, usage by feature, model, group and assistant, export, and privacy controls (aggregation, pseudonymisation, works-council-friendly settings). Distinguish aggregate adoption metrics from per-person monitoring.

Unique AI

Agent cross-checked Support: partial

Analytics exports cover adoption, assistants, models and tools with privacy controls.

Limitations

Exact group breakdowns, every dashboard entitlement and suppression beyond exported identifiers need verification.

adoption_dashboard Is there an adoption dashboard? [supported]: Reports cover daily/weekly/monthly active users, assistant interactions and document reference statistics. Sources: unique-ai:unique-ai:analytics.
breakdowns Can usage be broken down by group, model, feature and assistant? [partial]: Detailed exports include assistant/model, tool calls, skills, cost and tokens. Group and cost-centre breakdowns were not established. Sources: unique-ai:unique-ai:analytics.
export Can usage data be exported? [supported]: Admins request and download reports with date/assistant filtering; model usage supports CSV. Sources: unique-ai:unique-ai:analytics, unique-ai:unique-ai:cost.
privacy_modes Can per-person reporting be limited or pseudonymised? [supported]: Environment flags anonymise export emails, optionally with stable pseudonyms; detailed exports respect organisation settings. Sources: unique-ai:unique-ai:analytics.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:analytics, unique-ai:unique-ai:cost.

ADM-05 Spend controls and cost allocation

Judgment rule: Record cost visibility per user, group and cost centre, budgets and their behaviour (alert, block, downgrade model), pooled credits, and chargeback exports. Test a small budget on a trial where possible. Distinguish billed amounts, estimates and unavailable amounts.

Unique AI

Agent cross-checked Support: partial

Off-by-default beta cost management offers estimated spend and enforced daily caps.

Limitations

No small-budget fixture was run. Estimates are configurable cost-sheet values, not a public billing tariff; cost-centre and full invoice reconciliation remain gaps.

cost_visibility Can admins see cost or consumption per user and group? [partial]: Tracks estimated USD/input/output tokens by user, model, assistant/app; a billed invoice reconciliation or group cost view is not established. Sources: unique-ai:unique-ai:cost.
budgets Can budgets be set per user or group? [supported]: Company fallback, group and user daily USD caps; user override first, highest applicable group limit next. Sources: unique-ai:unique-ai:cost.
budget_enforcement Does reaching a budget alert, block or downgrade? [supported]: Before the next call, accumulated daily spend at cap blocks the model call; resets at local midnight. Docs warn caps may break background-job SLAs. Sources: unique-ai:unique-ai:cost.
chargeback Can costs be exported per cost centre? [partial]: CSV exports contain user/app/assistant/model/token/cost metadata, with privacy controls. Cost-centre attribution is not documented. Sources: unique-ai:unique-ai:cost.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:cost.

ADM-06 Audit logs and compliance export

Judgment rule: Record which events are logged (logins, admin changes, prompts, responses, file access, connector queries), retention, export or SIEM streaming, and eDiscovery/legal hold. Use the word audit only with a defined retention and integrity model.

Unique AI

Agent cross-checked Support: partial

API audit events, detailed interaction exports and Single-Tenant security-event streaming are documented. A single authoritative retention period for the customer's audit records is not established.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; Audit Logs, Analytics and Single-Tenant Security Events Export availability and deployment/provider configuration depend on the customer agreement; add_ons=SEE for the contracted Single-Tenant deployment; deployment_mode=not specified; region=CH/EU/US/UK/APAC per deployment; inference and external tools require separate region configuration; os=not specified.

Limitations

API security-event records, detailed prompt/answer interaction exports, SEE replay and generic operational logging are separate sources/data flows. Their public retention figures are not substituted for each other.
No executed customer retention override or legal-hold/eDiscovery fixture was inspected.

admin_events Are admin and security events logged? [supported]: Event logging covers privileged public/internal API interactions; many self-identity/chat operations are excluded. Basic mode is default; verbose content mode is opt-in. Sources: unique-ai:unique-ai:audit.
content_logging Can prompts and answers be retained for compliance, and is it configurable? [supported]: Detailed chat interaction exports contain prompts/answers for authorised admins and respect configured privacy. Verbose event logging can include CID but is discouraged. Sources: unique-ai:unique-ai:analytics, unique-ai:unique-ai:audit.
siem_export Can logs stream to a SIEM or be exported? [supported]: SEE offers Single-Tenant Azure Defender and Activity events via Event Hub to Sentinel/Splunk, seven-day replay. It excludes tenant-level Entra events and is not a prompt-content audit stream. Sources: unique-ai:unique-ai:security-export.
ediscovery Are eDiscovery and legal hold supported? [unknown]: No public legal-hold/eDiscovery workflow or complete event-log immutability design was established.
log_retention How long are logs kept? [unknown]: Different sources describe different record scopes. The API event guide gives basic-mode MT events 90 days and ST events five years, including post-termination ST retention. DPA 3.15 says audit logs ten years without mapping its audit-log category to those API events. The generic hosted policy says logging up to one year after termination or longer if legally required. SEE has a separate seven-day Event Hub replay window, not the lifetime of all audit records. No customer agreement or authoritative record-class/precedence mapping was retrieved; the applicable audit retention period is unknown.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-audit, unique-ai:unique-ai:analytics, unique-ai:unique-ai:security-export, unique-ai:unique-ai:review2-dpa, unique-ai:unique-ai:review2-retention.

ADM-07 Data loss prevention and personal data

Judgment rule: Record detection and handling of personal and confidential data before it reaches a model: redaction, masking, blocking, warning. Record integration with Purview, Netskope or similar, and whether filtering runs locally or at a third party. Send a fixture prompt with a Swiss AHV number and IBAN and record the outcome.

Unique AI

Agent cross-checked Support: partial

DLP is primarily customer infrastructure and integration; it is not a verified default native redaction feature.

Limitations

Default prompt/file PII detection, specific redaction/block behaviour, custom patterns, supported Purview/Netskope policies and native AHV/IBAN handling were not established or tested. Data redaction docs concern retention, not PII masking.

pii_detection Is personal data detected in prompts and files? [unknown]: Is personal data detected in prompts and files? Unknown. Default prompt/file PII detection, specific redaction/block behaviour, custom patterns, supported Purview/Netskope policies and native AHV/IBAN handling were not established or tested. Data redaction docs concern retention, not PII masking.
pii_action Is it redacted, masked, blocked or only warned? [unknown]: Is it redacted, masked, blocked or only warned? Unknown. Default prompt/file PII detection, specific redaction/block behaviour, custom patterns, supported Purview/Netskope policies and native AHV/IBAN handling were not established or tested. Data redaction docs concern retention, not PII masking.
custom_patterns Can admins add custom patterns or terms? [unknown]: Can admins add custom patterns or terms? Unknown. Default prompt/file PII detection, specific redaction/block behaviour, custom patterns, supported Purview/Netskope policies and native AHV/IBAN handling were not established or tested. Data redaction docs concern retention, not PII masking.
dlp_integration Does it integrate with an existing DLP product? [partial]: Documents customer HTTPS proxy/interceptor DLP and post-chat Analytics API scanning; Compliance Layer says pre-LLM calls can be implemented. Multi-Tenant matrix explicitly excludes DLP integration. Sources: unique-ai:unique-ai:dlp, unique-ai:unique-ai:compliance.
filter_location Where does filtering run? [partial]: Proxy inspection runs in customer egress infrastructure; Analytics API supports customer post-chat inspection. A deployment-specific pre-LLM service may be implemented. Sources: unique-ai:unique-ai:dlp, unique-ai:unique-ai:compliance.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:dlp, unique-ai:unique-ai:compliance.

ADM-08 AI policy and guardrails

Judgment rule: Record organisation-wide system instructions, acceptable-use notices and acknowledgement, content filters, topic restrictions, AI-output labelling, and admin visibility of policy violations. Separate enforced policy from instructions the model is expected to follow.

Unique AI

Agent cross-checked Support: partial

Space instructions/banners and customer-specific login terms are documented.

Limitations

Org-wide instruction enforcement, customer content-filter policy controls and administrator violation reports are not established. Prompt guidance is not guaranteed enforcement.

org_instructions Can admins set organisation-wide instructions? [partial]: Admins set Space-level system instructions and persistent banners. A mandatory organisation-wide instruction hierarchy is not specified. Sources: unique-ai:unique-ai:agent.
usage_notice Can users be required to accept a usage policy? [supported]: First login requires terms acceptance; custom terms/editing/deactivation depend on MSA eligibility and admin GraphQL role. Acceptance/version data can be exported. Sources: unique-ai:unique-ai:policy.
content_filters Are content filters configurable? [partial]: Single-tenant Azure prompts pass through Azure content filters; Compliance matrix marks hosted modes with content filtering. Customer filter configuration and scope remain unspecified. Sources: unique-ai:unique-ai:single, unique-ai:unique-ai:compliance.
violation_reporting Are policy hits visible to admins? [unknown]: Are policy hits visible to admins? Unknown. Org-wide instruction enforcement, customer content-filter policy controls and administrator violation reports are not established. Prompt guidance is not guaranteed enforcement.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:agent, unique-ai:unique-ai:policy, unique-ai:unique-ai:single, unique-ai:unique-ai:compliance.

Data protection and sovereignty

DATA-01 Hosting location and data residency

Judgment rule: Record country or region separately for chat storage, files and indexes, model inference, backups, logs and telemetry, and support access. Name the cloud provider. Record whether residency is default, a paid option or plan-dependent. "EU data boundary" programmes are recorded with their documented exceptions.

Unique AI

Agent cross-checked Support: partial

Documentation offers CH North multitenant hosting and selectable single/customer-hosted regions, with model processing configured separately. The customer's binding residency schedule and full backup/log/support map remain unknown.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; hosted or customer-hosted platform; separately selected model/tool routes availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=Documented options: CH North hosted MT; selected ST/customer region; provider/SKU/tool inference region separately configured; actual contracted boundaries unverified; os=not specified.

Limitations

Available architecture/region options are not a signed residency schedule. Storage, model inference, backups/logs, support access and external tools must be scoped separately.
DPA CH-storage and chosen-location statements are not reconciled by silently treating one as the current customer contract.

storage_region Where are chats, files and indexes stored? [partial]: The deployment matrix documents Azure Switzerland North for hosted multitenant and chosen supported regions for Single Tenant. Tenant storage includes PostgreSQL, documents and vector data. These are configuration options; no customer-specific storage schedule was inspected. Sources: unique-ai:unique-ai:review2-compliance, unique-ai:unique-ai:review2-keys.
inference_region Where do model calls run, per provider? [partial]: The model catalogue separates Azure regional/Data Zone/Global SKUs and external provider routes, with some Swiss deployments and some global-only services. Actual selected provider/SKU/tool routing must be configured for the contracted deployment. CH storage is not a verified CH inference guarantee. Sources: unique-ai:unique-ai:review2-models, unique-ai:unique-ai:deepresearch.
backups_logs_region Where are backups, logs and telemetry? [partial]: Single-tenant backup resources follow the configured landing-zone design and SEE sends events to a separately selected consumer region. An exhaustive physical replica/backup/log/telemetry country matrix was not retrieved. Sources: unique-ai:unique-ai:review2-single, unique-ai:unique-ai:security-export.
support_access_region From where can vendor staff access data? [unknown]: The public DPA gives need-to-know access controls and customer-controlled on-premises support, but not an exhaustive physical country list for vendor/subprocessor support access. Office addresses are not used as support-access locations.
swiss_hosting Is Swiss hosting available? [partial]: CH North is a documented hosted-MT option, and selected-region ST can be Swiss. This records available deployment options, not the customer's end-to-end storage/inference/support contract. Sources: unique-ai:unique-ai:review2-compliance, unique-ai:unique-ai:review2-models.
residency_plan Which plan or option provides residency? [unknown]: DPA 6.1 says storage stays in Switzerland, while its Appendix A allows a client-chosen Microsoft location and current platform docs offer selected jurisdictions. No current customer MSA, precedence clause or deployment schedule was inspected to resolve the binding scope. A contractual CH-wide residency commitment is not asserted.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-compliance, unique-ai:unique-ai:review2-models, unique-ai:unique-ai:review2-keys, unique-ai:unique-ai:review2-single, unique-ai:unique-ai:review2-dpa, unique-ai:unique-ai:security-export, unique-ai:unique-ai:deepresearch.

DATA-02 Deployment options

Judgment rule: Record multi-tenant SaaS, dedicated single-tenant instance, deployment in the customer's cloud account, on-premise and air-gapped as separate options with plan and availability. Record which features are missing in each mode, who operates it, and the documented seat or contract threshold for each mode, which can differ from the enterprise plan threshold (e.g. dedicated SaaS from 2,000 seats, own cloud from 5,000). A local-network connector or gateway that reaches on-prem data from a vendor-hosted platform is KNOW-03, not on-premise deployment.

Unique AI

Agent cross-checked Support: partial

Current docs offer shared SaaS, dedicated instances, customer cloud and customer-operated on-premise/air-gapped modes.

Limitations

No deployed tenant was inspected. Seat/contract thresholds and an exhaustive offline feature-parity matrix remain unknown; docs establish an option, not universal entitlement.

multi_tenant_saas Is a shared SaaS offered? [supported]: Installation overview lists Multi Tenant on Unique cloud. Sources: unique-ai:unique-ai:install.
dedicated_instance Is a dedicated single-tenant instance offered, from what size? [partial]: Unique-managed Single Tenant has its own Azure landing zone separated from other customers. Minimum seats/contract value are unpublished. Sources: unique-ai:unique-ai:single.
customer_cloud Can it run in the customer's own cloud account, from what size? [partial]: Operator guide documents deployment in customer cloud account with Unique software/support; cloud-specific rollout is contracted. Sources: unique-ai:unique-ai:operators.
on_premise Can the whole platform run on customer infrastructure, from what size? [partial]: Customer provisions/manages infrastructure on own premises, may run own models and can exclude Unique access. Minimum commitment not published. Sources: unique-ai:unique-ai:onpremsecurity.
air_gapped Can it run without internet access? [partial]: IT-operator overview explicitly offers a full air-gapped on-premise deployment. Compatibility and feature limitations depend on customer setup. Sources: unique-ai:unique-ai:operators, unique-ai:unique-ai:onpremsecurity.
mode_feature_gaps Which features are missing outside SaaS? [partial]: Compliance matrix separates SSO/content filter/DLP/PAM/key capabilities by mode. Azure Chat Code Execution/voice and external web/providers require those services and cannot be assumed in an isolated environment. Sources: unique-ai:unique-ai:compliance, unique-ai:unique-ai:code, unique-ai:unique-ai:voice.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:install, unique-ai:unique-ai:single, unique-ai:unique-ai:operators, unique-ai:unique-ai:onpremsecurity, unique-ai:unique-ai:compliance, unique-ai:unique-ai:code, unique-ai:unique-ai:voice.

DATA-03 Training use and provider retention

Judgment rule: Record whether customer content trains vendor or provider models, by plan and default; zero-data-retention arrangements with model providers; abuse-monitoring retention and human review; and the model provider's terms that apply. A "no training" claim on a marketing page is `vendor-claimed` until the contract or DPA says so.

Unique AI

Agent cross-checked Support: partial

Hosted Azure documentation says no model training and disabled Microsoft prompt storage/human review; other provider routes require separate terms.

Limitations

No signed Azure amendment was inspected. Retention and review conditions for Anthropic, Gemini, Together, xAI and direct OpenAI research remain unverified; do not generalise the Azure-only policy.

vendor_training Does the vendor train on customer content? [supported]: Hosted retention policy says no customer data is used for AI/neural-network training. Public terms allow service development and anonymised metadata use; that is not a blanket prohibition of all secondary processing. Sources: unique-ai:unique-ai:retention, unique-ai:unique-ai:terms.
provider_training Do model providers train on it under this product's contracts? [partial]: Hosted Azure deployment is separate from public OpenAI and training is excluded. Other catalogue providers/direct OpenAI research need their applicable customer terms. Sources: unique-ai:unique-ai:retention, unique-ai:unique-ai:models, unique-ai:unique-ai:deepresearch.
provider_retention How long do model providers keep prompts, including for abuse monitoring? [supported]: For hosted Azure OpenAI, documents/prompts/answers process in memory only, with logging/human-review opt-out. Ordinary application chats persist in Unique databases. Sources: unique-ai:unique-ai:retention.
zero_retention Is zero data retention available, and on which plan? [partial]: Hosted Azure model calls use no provider content storage per opt-out; this is not zero retention of Unique application history, logs or other provider routes. Sources: unique-ai:unique-ai:retention.
human_review Can provider staff review content? [partial]: Hosted Azure guide says Microsoft prompt/response human review is disabled. Vendor staff need-to-know access and alternate model-provider review conditions are separate. Sources: unique-ai:unique-ai:retention, unique-ai:unique-ai:dpa.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:retention, unique-ai:unique-ai:terms, unique-ai:unique-ai:models, unique-ai:unique-ai:deepresearch, unique-ai:unique-ai:dpa.

DATA-04 Retention and deletion

Judgment rule: Record default retention for chats, files, indexes and logs; admin-configured retention; user deletion; deletion after contract end; and deletion confirmation.

Unique AI

Agent cross-checked Support: partial

Chat content clearing, chat uploads, KB documents, contract-end content deletion, recoverable files, backups and audit records have separate policies. No common purge deadline is established for all copies.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; Chat, Knowledge Base and hosted retention configuration; customer-hosted policy separately controlled availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=Contracted storage/provider region; hosted-policy timings not extended to every customer-hosted system; os=not specified.

Limitations

Hosted content policy expressly permits contract-specific variation. Its timings do not define API/DPA audit-record retention.
No tenant configuration, customer MSA, final backup purge or external-provider deletion fixture was inspected.

default_retention What is retained and for how long by default? [partial]: The chat-upload guide describes an admin-configured expiry, typically two weeks, for chat uploads. Hosted policy lists no minimum for KB/blob/vector/database content. Neither figure defines audit-event retention or proves a tenant's configured default. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:review2-retention.
admin_retention Can admins set retention periods? [supported]: DATA_RETENTION_IN_DAYS drives the daily midnight chat-content clearing job; unset/zero disables it. This clears the content categories named by the redaction guide, not all uploads, KB, external tool records or audit events. Sources: unique-ai:unique-ai:redaction.
user_deletion Can users delete their chats and files permanently? [partial]: The redaction guide says deleting a chat removes its Unique chat-content copy and marks the chat record deleted. Separate upload, KB, external-tool, soft-delete, backup and audit policies are not collapsed into that action. Sources: unique-ai:unique-ai:redaction.
contract_end_deletion What is deleted at contract end, and when? [partial]: For Unique-hosted ordinary content, the published policy states a standard 30-day post-termination period unless the customer contract differs. It separately describes recoverable files for 30 days after deletion and database backups for two further weeks. DPA 6.1 requires return/deletion subject to legal record-keeping. Audit-log classes and their contractual lifetime remain unknown under ADM-06; this is not a deadline for deletion of every copy. Sources: unique-ai:unique-ai:review2-retention, unique-ai:unique-ai:review2-dpa.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:files, unique-ai:unique-ai:redaction, unique-ai:unique-ai:review2-retention, unique-ai:unique-ai:review2-dpa.

DATA-05 Sub-processors and foreign-law exposure

Judgment rule: Record the published sub-processor list with countries, change-notification rules, the vendor's and parent company's jurisdiction, transfer mechanisms (EU-US Data Privacy Framework, Swiss-US DPF, SCCs) and the vendor's documented position on foreign government access requests (e.g. US CLOUD Act). Record facts, not a legal assessment.

Unique AI

Agent cross-checked Support: partial

The dated public DPA lists Microsoft, but a complete applicable current subprocessor/country schedule for all contracted model/tool routes was not retrieved. A technical provider catalogue is not substituted for that legal schedule.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; current platform infrastructure and enabled model/tool integrations availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=Actual customer provider/country/transfer schedule unverified; os=not specified.

Limitations

The provider catalogue establishes technical choices, not each party's processor role or the legal schedule for a customer contract.
No sales contact, executed current MSA or nonpublic schedule was obtained.

subprocessor_list Is a dated sub-processor list published? [unknown]: The July 2025 public DPA Appendix A lists Microsoft for infrastructure, Cognitive/Azure OpenAI and email. The current technical catalogue offers other provider/tool routes, which may have different contractual roles. No applicable complete current legal schedule or executed customer supplement was retrieved; the Microsoft-only list is not presented as complete.
subprocessor_countries Which countries do sub-processors operate in? [unknown]: DPA Appendix A names CH or a client-chosen Microsoft location; technical model routes separately include regional and global processing. A complete current legal provider/country and support-access schedule for the chosen deployment was not retrieved.
change_notification Are customers notified of sub-processor changes? [unknown]: No applicable current processor-change notification/objection clause for the customer's full provider schedule was retrieved from the DPA or current contract-access documentation.
parent_jurisdiction Which jurisdiction governs the vendor and its parent? [partial]: The current contact page identifies Unique AI AG in Switzerland and regional entities. The public DPA uses Unique AG. Ultimate parent/beneficial-owner jurisdiction and a customer-specific contracting entity were not independently established. Sources: unique-ai:unique-ai:contact, unique-ai:unique-ai:review2-dpa.
transfer_mechanism Which transfer mechanism applies to non-EU/CH processing? [unknown]: No complete customer-content transfer-instrument mapping per non-CH/EU provider was retrieved. The website visitor privacy policy is not substituted for customer-content terms.
government_access_policy Is there a published government-access or transparency policy? [unknown]: No current customer-content government-request/transparency policy was retrieved for the complete contracted provider chain. General banking/DPA assurances do not establish such a policy.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-dpa, unique-ai:unique-ai:review2-models, unique-ai:unique-ai:review2-governance, unique-ai:unique-ai:contact.

DATA-06 Contracts and regulatory coverage

Judgment rule: Record DPA availability and form (click-through, negotiable), reference to Swiss revDSG and GDPR, governing law and court, EU AI Act documentation (provider or deployer role, transparency materials), and sector-specific materials (FINMA outsourcing, banking secrecy, public sector). No compliance verdict.

Unique AI

Agent cross-checked Support: partial

A public DPA and finance-specific compliance material are available; actual contractual amendments were not inspected.

Limitations

DPA negotiability and signed sector amendments, product-specific EU AI Act roles and exact sector assurance scope remain unverified. This records materials, not a compliance verdict.

dpa Is a DPA available, and is it negotiable? [partial]: DPA accepts click-through/use or signature and defines Unique as processor for hosted MT/ST. Negotiability is not specified. Sources: unique-ai:unique-ai:dpa.
revdsg_gdpr Does the DPA cover revDSG and GDPR? [partial]: DPA addresses GDPR, Swiss FDPIC oversight and MSA-defined laws; Compliance Layer identifies nFDAP/Swiss M329 amendment. Exact signed revDSG/GDPR scope requires MSA. Sources: unique-ai:unique-ai:dpa, unique-ai:unique-ai:compliance.
governing_law Which law and court govern the contract? [supported]: DPA 7 specifies Swiss substantive law and Zürich 1 courts unless mandatory GDPR/MSA otherwise. Public ToS similarly names Swiss law. Sources: unique-ai:unique-ai:dpa, unique-ai:unique-ai:terms.
ai_act_materials Is EU AI Act documentation provided? [partial]: Trust centre describes per-use-case conformity assessments; ISO42001 and AI governance material are public. A complete provider/deployer AI Act technical file is not established. Sources: unique-ai:unique-ai:trust, unique-ai:unique-ai:compliance.
sector_materials Are FINMA or other sector materials provided? [partial]: Compliance Layer names Microsoft M453 FINMA audit-rights amendment, M744 professional/banking secrecy and M329 CH-DPA for Unique-hosted deployments. Actual annexes and current signed scope are not public. Sources: unique-ai:unique-ai:compliance.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:dpa, unique-ai:unique-ai:compliance, unique-ai:unique-ai:terms, unique-ai:unique-ai:trust.

DATA-07 Certifications and attestations

Judgment rule: Record each certificate or report separately: ISO 27001, 27017, 27018, 27701, 42001; SOC 2 Type I/II with period; BSI C5; TISAX; national schemes. Record issuer, scope (which product and entity), validity and access route. A cloud provider's certificate does not certify the product.

Unique AI

Agent cross-checked Support: partial

New public 2026 ATTESTA renewals establish ISO27001 and ISO9001 validity to 31 May 2029; the public TÜV SÜD ISO42001 certificate runs to 6 March 2028. SOC2 Type2 is vendor-documented, with full report evidence NDA-gated.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=ATTESTA 10 June 2026 renewals; TÜV SÜD 7 March 2025 certificate; plan=Contracted Unique AI platform; platform assurance; certificates cover the named legal entity and stated scopes, not every module/provider availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=Certificates name Unique AG, Zurich, Switzerland; module/provider coverage depends on the stated audit scope; os=not specified.

Limitations

Current ISO27001/9001 evidence is the 2026 renewal plus matching register entries; the stale website-linked PDF is not used to deny current certification.
SOC chronology and access process are vendor-documented, not a reviewed independent SOC report. Certificate scopes are entity-specific and not a universal module/provider assurance.

iso_27001 ISO 27001 with scope and validity? [supported]: Inspected the 2026 ATTESTA Schweizer Zertifizierungsgesellschaft AG ISO27001:2022 certificate for Unique AG, registration 223-06-018.I. It covers software development/operation/trading/distribution and IT services, is issued 10 June 2026 after 4 June surveillance and expires 31 May 2029. The certificate-linked public register matches the number and expiry. The marketing-linked 2025 PDF is an older artifact, not the current renewal. Sources: unique-ai:unique-ai:review2-iso27001, unique-ai:unique-ai:review2-iso-register, unique-ai:unique-ai:review2-iso-docs.
soc2 SOC 2 type, period and covered service? [partial]: The dedicated SOC page documents Type1 in December 2023 followed by Type2 in January 2025. This explains the historical Type1 references rather than proving an ongoing Type1-only state. Current Type2 status is a vendor claim; the auditor, actual report period and service/control scope were not inspected because the full report is NDA-gated. Sources: unique-ai:unique-ai:review2-soc.
c5 BSI C5 attestation? [unknown]: No public BSI C5 attestation was retrieved from the certification pages/trust centre or targeted first-party search.
other_certifications Which other certificates (27701, 42001, TISAX)? [partial]: Inspected ATTESTA ISO9001:2015 renewal 223-06-017.Q for the same Unique AG software/IT scope, expiring 31 May 2029, matched in the public register. Separately inspected TÜV SÜD South Asia ISO42001:2023 certificate 99 420 00001, valid 7 March 2025 to 6 March 2028, for development/deployment of financial-services SaaS using foundation or client models. Current full ISAE/FINMA report scope was not inspected. Sources: unique-ai:unique-ai:review2-iso9001, unique-ai:unique-ai:review2-iso-register, unique-ai:unique-ai:iso42001.
report_access How can customers obtain reports? [supported]: The ISO documentation provides public certificate downloads. It says the ISO27001 Statement of Applicability requires a valid NDA. The dedicated SOC page likewise makes the audit report/certificate available on request after NDA. No NDA or sales request was made. Sources: unique-ai:unique-ai:review2-iso-docs, unique-ai:unique-ai:review2-soc.
pentest Are independent penetration test summaries available? [partial]: The public DPA describes internal/external security testing and the trust page describes tests/bug bounty. An independent penetration-test report, current scope or summary was not inspected. Sources: unique-ai:unique-ai:review2-dpa, unique-ai:unique-ai:trust.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-iso27001, unique-ai:unique-ai:review2-iso9001, unique-ai:unique-ai:review2-iso-register, unique-ai:unique-ai:review2-iso-docs, unique-ai:unique-ai:iso42001, unique-ai:unique-ai:review2-soc, unique-ai:unique-ai:review2-dpa, unique-ai:unique-ai:trust.

DATA-08 Encryption and key control

Judgment rule: Record encryption at rest and in transit, customer-managed keys (BYOK/HYOK), key location, and private networking (private link, IP allow-lists, VPN).

Unique AI

Agent cross-checked Support: partial

Deployment guides document encryption, isolated Single Tenant resources and optional IP filtering. Exact current MT PMK/CMK/HSM configuration and key sizes are not established across the conflicting pages.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; hosted MT/ST encryption and optional customer-controlled self-hosting availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=Contracted Azure tenant/storage/backup region; actual key-vault/HSM location unverified; os=not specified.

Limitations

No exact RSA/HSM key size or actual MT PMK/CMK design is asserted. A customer-specific deployment/key diagram and key-vault region were not inspected.
Customer-generated imported keys held by Unique do not establish customer-exclusive key custody; full control is a separate self-hosting option.

encryption_at_rest Is data encrypted at rest? [partial]: The compliance matrix documents encryption at rest and in transit across deployment options, with AES256/TLS1.2+ in its controls text. No exact current RSA key size is asserted: the ST architecture says 2048-bit RSA-HSM while the hosted retention page says 4096-bit. An implementation-specific key design was not retrieved. Sources: unique-ai:unique-ai:review2-compliance, unique-ai:unique-ai:review2-single, unique-ai:unique-ai:review2-retention.
customer_managed_keys Can the customer manage or hold the keys? [partial]: The infrastructure key guide allows a customer-generated key to be imported for a Unique-hosted Single Tenant, with Unique holding/managing it. It directs clients needing exclusive control of all storage keys to customer self-hosting. This is not HYOK or exclusive customer custody in Unique hosting. Actual MT PMK/CMK/HSM configuration is unknown because the key guide and hosted retention matrix disagree. Sources: unique-ai:unique-ai:review2-keys, unique-ai:unique-ai:review2-retention.
private_networking Is private connectivity or IP restriction available? [partial]: The Single-Tenant WAF supports optional IP blocking. The separate SEE consumer endpoint is public TLS with optional allow-list and no standard private endpoint/peering. These are distinct scopes; other private-link options for the customer deployment were not verified. Sources: unique-ai:unique-ai:review2-single, unique-ai:unique-ai:security-export.
tenant_isolation How are customers separated in shared SaaS? [supported]: The Single-Tenant architecture describes a separate customer landing zone under Unique's subscription, isolating its storage/processing from other customers. This documents that ST architecture, not the isolation configuration of every hosted or customer-managed instance. Sources: unique-ai:unique-ai:review2-single.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-compliance, unique-ai:unique-ai:review2-single, unique-ai:unique-ai:review2-retention, unique-ai:unique-ai:review2-keys, unique-ai:unique-ai:security-export.

DATA-09 Professional secrecy

Judgment rule: For professional secrecy holders (Berufsgeheimnisträger: lawyers, doctors, auditors, tax advisers) under § 203 StGB (DE) and Art. 321 StGB (CH), record whether the vendor signs a contractual secrecy undertaking or a § 203 agreement that binds its staff and its sub-processors; whether this comes as a dedicated offering or a paid add-on, and on which plan; who can access content in plain text (vendor staff, sub-processors, model providers); and the documentation provided. A DPA or a no-training clause alone is not a secrecy undertaking. Banking secrecy and FINMA outsourcing belong to DATA-06, not here. Record contract facts, not a legal assessment.

Unique AI

Agent cross-checked Support: partial

Unique documents general staff/external-party access controls and customer-controlled on-premises access. A professional-secrecy undertaking under § 203 StGB or Art. 321 StGB, its subprocessor flow-down and qualifying plan/add-on remain unverified. Financial-sector banking-secrecy amendments are recorded separately under DATA-06.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI; profession-specific eligibility and fees unpublished; add_ons=none stated; deployment_mode=not specified; region=Contracted hosting jurisdiction; customer-controlled on-premises access described separately from Unique-hosted access; os=not specified.

Limitations

No executed professional-secrecy undertaking or § 203 / Art. 321 contract was inspected. General DPA confidentiality is not treated as such an undertaking.
The M744 professional-secrecy wording appears in a banking-secrecy/financial-services framework; no inference is made to lawyers, doctors, auditors or tax advisers.
Customer-controlled on-premises access does not establish the same access conditions for Unique-hosted deployments or all model-provider routes. Full plaintext-access mapping remains unknown.

secrecy_undertaking Does the vendor sign a secrecy undertaking or § 203 StGB / Art. 321 StGB agreement covering its own staff? [unknown]: No public Unique AI-specific § 203 StGB or Art. 321 StGB secrecy undertaking covering staff was located in the DPA, compliance documentation or targeted first-party search on 21 October 2026. The DPA confidentiality commitment alone does not establish a professional-secrecy undertaking. No executed or negotiated customer agreement was inspected.
subprocessor_coverage Does the undertaking extend to its sub-processors, and is that documented? [unknown]: The DPA says employees, contractual subprocessors and other service providers are bound to confidentiality of sensitive information. A professional-secrecy-specific obligation extending through the subprocessor/model-provider chain was not verified. The Microsoft M744 banking-secrecy amendment mentioned in the financial-services framework is not counted as evidence of § 203 / Art. 321 coverage for professional secrecy holders.
secrecy_offering Is it a dedicated offering or add-on, and on which plan? [unknown]: No qualifying professional-secrecy plan, dedicated offering, add-on tariff or named eligibility rule was located. The documented financial-services amendments and general deployment choices do not establish a separate § 203 / Art. 321 offering. Contract-specific terms and fees remain unknown.
content_access Who can access content in plain text: vendor staff, sub-processors, model providers? [partial]: DPA Appendix B permits employees or external parties access to personal/confidential information on a need-to-know basis. For on-premises support, customers control remote connections and Unique staff cannot access a customer system without customer knowledge or full active participation. The on-premises architecture also allows operation without any Unique access under customer control. These sources do not exhaustively identify plaintext access at each hosted subprocessor/model provider or profession-specific access conditions. Sources: unique-ai:unique-ai:dpa, unique-ai:unique-ai:onpremsecurity.
secrecy_documentation Is the offering publicly documented (contract template, data-flow description, professional-body guidance)? [unknown]: A public DPA and on-premises architecture describe confidentiality and access controls, but no profession-specific secrecy template or professional-body guidance was located. The compliance page describes Microsoft M453/M744/M329 in a financial-services context; those banking/FINMA disclosures belong to DATA-06 and do not establish the missing professional-secrecy contract.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:dpa, unique-ai:unique-ai:onpremsecurity, unique-ai:unique-ai:compliance.

Commercials and vendor

VEN-01 Pricing model and list prices

Judgment rule: Record pricing unit (per seat, per active user, usage-based, platform fee, hybrid), list prices per plan in original currency with date, included usage and credits, add-on prices (connectors, agents, image generation) and overage rates. Cite public prices or dated quotes. Per-seat versus usage is a fact, not a quality judgment.

Unique AI

Agent cross-checked Support: partial

Applicable current platform pricing unit, list tariff, included usage and overages remain unknown. A separately documented Single-Tenant SEE consumption/engineering charge is retained; legacy sales-conversation terms are excluded.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; platform subscription and separately enabled Single-Tenant SEE availability and deployment/provider configuration depend on the customer agreement; add_ons=SEE scope only where contracted; deployment_mode=not specified; region=Vendor currency and applicable invoicing terms not established for the platform; SEE guide separately quotes indicative CHF consumption; os=not specified.

Limitations

The SEE cost finding is scoped only to that Single-Tenant integration. It does not price the platform subscription or all modules.
The 2023 sales-conversation ToS is excluded as evidence of current Unique AI commercial terms; no sales quote or private contract was requested.

pricing_unit What is the pricing unit? [unknown]: No applicable current platform MSA/order form or tariff establishing seat, active-user, platform or usage pricing units was retrieved. The public 18 October 2023 ToS defines a sales-conversation intelligence service, so its subscription/Order Form clauses are not applied to current Unique AI.
list_prices What are list prices per plan, with date and currency? [unknown]: No applicable current platform list tariff, tier names or primary invoicing currency was retrieved from platform/contract documentation or targeted first-party search.
included_usage What usage is included? [unknown]: No applicable current platform agreement specifying included model tokens, storage or module allowances was publicly retrieved.
add_ons Which features cost extra? [partial]: Single-Tenant SEE passes through Azure consumption without markup; guide observes CHF20-50/month/environment, not a fixed quote, with initial engineering at agreed rates. Other module charges are unpublished. Sources: unique-ai:unique-ai:security-export.
overage What do usage beyond limits and premium models cost? [unknown]: No applicable current platform overage or premium-model sales tariff was retrieved. Cost-management model estimates are not substituted for a customer sales tariff.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:security-export, unique-ai:unique-ai:review2-governance, unique-ai:unique-ai:review2-terms-admin.

VEN-02 Scenario cost

Judgment rule: Price three workloads for 12 months: 200, 1,000 and 5,000 users with a disclosed light/regular/power usage mix, in CHF and EUR excluding VAT. Include licences, platform fees, add-ons, model usage and, for self-hosted options, a stated operating assumption. Separate fixed, variable, included and unknown costs. No invented discounts.

Unique AI

Agent cross-checked Support: partial

ALM-14 shared cost assumptions v1.4, 2026-10-21. Public enterprise pricing is insufficient for the required 200/1,000/5,000-user annual totals.

Limitations

ALM-14 shared cost assumptions v1.4, 2026-10-21. Horizon 12 months, list prices excluding VAT, no invented discounts. Mix 60% light at 100 messages/user/month, 30% regular at 400, 10% power at 1,500. Where model usage is separately billed, each message has 3,000 input and 600 output tokens, giving 990k input plus 198k output per user/month. At 200/1,000/5,000 users this is 198m/990m/4.95bn input and 39.6m/198m/990m output per month. All three seat counts require an unpublished enterprise/custom price; no quote was requested. CHF and EUR ex-VAT totals, fixed charges, variable charges and add-ons remain unknown. FX dated 2026-10-20: EUR 1 = CHF 0.9237 and USD 1.1215; CHF to EUR divide by 0.9237, USD to EUR divide by 1.1215, source ECB. No numeric conversion or proxy model tariff is applied because there is no applicable public enterprise base/usage rate.

cost_200 Year-one cost at 200 users? [unknown]: ALM-14 shared cost assumptions v1.4, 2026-10-21. Horizon 12 months, list prices excluding VAT, no invented discounts. Mix 60% light at 100 messages/user/month, 30% regular at 400, 10% power at 1,500. Where model usage is separately billed, each message has 3,000 input and 600 output tokens, giving 990k input plus 198k output per user/month. At 200/1,000/5,000 users this is 198m/990m/4.95bn input and 39.6m/198m/990m output per month. All three seat counts require an unpublished enterprise/custom price; no quote was requested. CHF and EUR ex-VAT totals, fixed charges, variable charges and add-ons remain unknown. FX dated 2026-10-20: EUR 1 = CHF 0.9237 and USD 1.1215; CHF to EUR divide by 0.9237, USD to EUR divide by 1.1215, source ECB. No numeric conversion or proxy model tariff is applied because there is no applicable public enterprise base/usage rate.
cost_1000 Year-one cost at 1,000 users? [unknown]: ALM-14 shared cost assumptions v1.4, 2026-10-21. Horizon 12 months, list prices excluding VAT, no invented discounts. Mix 60% light at 100 messages/user/month, 30% regular at 400, 10% power at 1,500. Where model usage is separately billed, each message has 3,000 input and 600 output tokens, giving 990k input plus 198k output per user/month. At 200/1,000/5,000 users this is 198m/990m/4.95bn input and 39.6m/198m/990m output per month. All three seat counts require an unpublished enterprise/custom price; no quote was requested. CHF and EUR ex-VAT totals, fixed charges, variable charges and add-ons remain unknown. FX dated 2026-10-20: EUR 1 = CHF 0.9237 and USD 1.1215; CHF to EUR divide by 0.9237, USD to EUR divide by 1.1215, source ECB. No numeric conversion or proxy model tariff is applied because there is no applicable public enterprise base/usage rate.
cost_5000 Year-one cost at 5,000 users? [unknown]: ALM-14 shared cost assumptions v1.4, 2026-10-21. Horizon 12 months, list prices excluding VAT, no invented discounts. Mix 60% light at 100 messages/user/month, 30% regular at 400, 10% power at 1,500. Where model usage is separately billed, each message has 3,000 input and 600 output tokens, giving 990k input plus 198k output per user/month. At 200/1,000/5,000 users this is 198m/990m/4.95bn input and 39.6m/198m/990m output per month. All three seat counts require an unpublished enterprise/custom price; no quote was requested. CHF and EUR ex-VAT totals, fixed charges, variable charges and add-ons remain unknown. FX dated 2026-10-20: EUR 1 = CHF 0.9237 and USD 1.1215; CHF to EUR divide by 0.9237, USD to EUR divide by 1.1215, source ECB. No numeric conversion or proxy model tariff is applied because there is no applicable public enterprise base/usage rate.
variable_share Which part depends on usage? [unknown]: ALM-14 shared cost assumptions v1.4, 2026-10-21. Horizon 12 months, list prices excluding VAT, no invented discounts. Mix 60% light at 100 messages/user/month, 30% regular at 400, 10% power at 1,500. Where model usage is separately billed, each message has 3,000 input and 600 output tokens, giving 990k input plus 198k output per user/month. At 200/1,000/5,000 users this is 198m/990m/4.95bn input and 39.6m/198m/990m output per month. All three seat counts require an unpublished enterprise/custom price; no quote was requested. CHF and EUR ex-VAT totals, fixed charges, variable charges and add-ons remain unknown. FX dated 2026-10-20: EUR 1 = CHF 0.9237 and USD 1.1215; CHF to EUR divide by 0.9237, USD to EUR divide by 1.1215, source ECB. No numeric conversion or proxy model tariff is applied because there is no applicable public enterprise base/usage rate.
unknown_parts Which costs could not be priced? [supported]: Subscription/order pricing, usage billing, provider contracts, deployment operation, connectors, execution, training and support charges are not publicly priced. SEE consumption example alone cannot price the platform. Sources: unique-ai:unique-ai:terms, unique-ai:unique-ai:platform, unique-ai:unique-ai:security-export.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:terms, unique-ai:unique-ai:platform, unique-ai:unique-ai:security-export, unique-ai:unique-ai:ecb-fx.

VEN-03 Evaluation access

Judgment rule: Distinguish free tier, self-serve trial, sales-led pilot, paid proof of concept and open-source self-install. Record card requirement, duration, seat limit, and which features (SSO, connectors, admin) the trial includes. Map which shared scenarios can be run without sales approval.

Unique AI

Agent cross-checked Support: partial

The public route is demo/evaluation by arrangement; self-service trial conditions were not established.

Limitations

No permanent business free tier or self-serve trial with duration/card/seat/SSO/connector/admin entitlements was located. No sales contact, login or fixture account was used.

free_tier Is there a free tier for business use? [unknown]: Is there a free tier for business use? Unknown. No permanent business free tier or self-serve trial with duration/card/seat/SSO/connector/admin entitlements was located. No sales contact, login or fixture account was used.
self_serve_trial Is a self-serve business trial available? [unknown]: Is a self-serve business trial available? Unknown. No permanent business free tier or self-serve trial with duration/card/seat/SSO/connector/admin entitlements was located. No sales contact, login or fixture account was used.
trial_scope Do trials include SSO, connectors and admin features? [unknown]: Do trials include SSO, connectors and admin features? Unknown. No permanent business free tier or self-serve trial with duration/card/seat/SSO/connector/admin entitlements was located. No sales contact, login or fixture account was used.
sales_gate Does meaningful evaluation require sales contact? [partial]: Current platform CTA requests a demo; public baseline ToS says Unique may offer evaluation at its discretion. This does not prove every meaningful evaluation is sales-gated. Sources: unique-ai:unique-ai:platform, unique-ai:unique-ai:terms.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:platform, unique-ai:unique-ai:terms.

VEN-04 Contract terms and pricing transparency

Judgment rule: Record minimum seats or commitment, contract length, monthly versus annual billing, price-change rules, cancellation and downgrade terms, and invoicing in CHF/EUR. Classify pricing as transparent, partially specified, quote-only or unknown, naming the missing fields. Quote-only pricing is a disclosed constraint.

Unique AI

Agent cross-checked Support: partial

Current platform minimums, term, cancellation and local invoicing remain unknown because the applicable agreement was not publicly retrieved. Current documentation makes legal contracts available on request; legacy sales-conversation renewal clauses are excluded.

Availability: generally-available · Delivery: native
Scope: surfaces=Unique AI Chat and administration, Unique API and Python SDK; version=not specified; plan=Contracted Unique AI platform; platform commercial agreement; actual agreement not inspected availability and deployment/provider configuration depend on the customer agreement; add_ons=none stated; deployment_mode=not specified; region=CHF/EUR invoicing and legal entity for a specific platform contract unverified; os=not specified.

Limitations

No current customer agreement or sales quote was requested. Contract availability on request does not establish price, term or cancellation conditions.

minimum_commitment What seat or spend minimum applies? [unknown]: No current platform seat/spend minimum was established from the public contract-access and platform pages.
contract_length What contract lengths are offered? [unknown]: No applicable current Unique AI agreement establishing subscription duration, renewal term or notice period was retrieved. The 2023 sales-conversation service term is not applied.
cancellation How can the contract be cancelled or reduced? [unknown]: Current Unique AI cancellation, refund, seat-reduction and nonrenewal rules were not retrieved. The 30-day renewal notice and quantity restrictions from the sales-conversation ToS are excluded.
local_invoicing Can it be invoiced in CHF or EUR from a European entity? [unknown]: CHF/EUR invoice currency, billing entity and local invoicing eligibility for a current platform contract were not established.
transparency_class Is pricing transparent, partial, quote-only or unknown? [partial]: The current governance framework says legal contracts and end-user T&Cs are available on request. No applicable public platform list tariff or complete commercial agreement was retrieved. This records an access/transparency gap, not a price or a customer-specific commitment. Sources: unique-ai:unique-ai:review2-governance, unique-ai:unique-ai:review2-terms-admin.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:review2-governance, unique-ai:unique-ai:review2-terms-admin.

VEN-05 Source licence

Judgment rule: Record licence per component (client, server, connectors, hosted extensions) and revision. Distinguish OSI open source, source-available with restrictions (branding clauses, user caps), and closed. Repository visibility is not a licence classification. Licences change between releases: record the current terms, and never carry a fork's historical licence over to the upstream product or the reverse.

Unique AI

Agent cross-checked Support: partial

The public reference repository uses restricted Unique License v1, while the Python SDK has an explicit MIT exception. Hosted platform component terms and enterprise-edition boundaries remain unverified.

Availability: source-only · Delivery: native
Scope: surfaces=Unique API and Python SDK, Unique AI Chat and administration; version=Unique-AG/ai commit afa596c9b4b0231895ff70f44029a9a525f746e9; plan=Public reference-code and Python SDK licences; current contracted Unique AI client/server/connector agreement remains unavailable; add_ons=none stated; deployment_mode=not specified; region=Public Unique License v1 specifies Swiss law and Zurich venue; this is not a hosted-service residency claim; os=not specified.

Limitations

The MIT exception applies to unique_sdk, not every directory or the hosted platform.
Root reference-code terms are not substituted for customer platform branding, seat or enterprise-edition conditions.
The 2023 sales-conversation service ToS remains excluded.

licence_type Which licence applies to each component? [partial]: The public Unique-AG/ai repository at afa596c9b4b0231895ff70f44029a9a525f746e9 defaults to Unique License v1, which restricts commercial use, distribution and modification without written approval. The unique_sdk directory is an explicit MIT exception, confirmed by LICENSE and package metadata. This is a component-level source-available/MIT distinction; current hosted client/server/connector customer terms and a full dependency inventory remain unknown. Sources: unique-ai:unique-ai:reference-licence-20261026, unique-ai:unique-ai:unique-licence-v1-20261026, unique-ai:unique-ai:sdk-licence-20261026, unique-ai:unique-ai:sdk-package-20261026.
branding_restrictions Do licence terms restrict branding or user counts? [partial]: Unique License v1 requires written approval to modify or redistribute covered reference code; the MIT SDK exception requires retention of copyright/licence notices. A current hosted-platform white-label/logo-removal obligation or a numerical user cap is not established by these public component licences. Sources: unique-ai:unique-ai:reference-licence-20261026, unique-ai:unique-ai:unique-licence-v1-20261026, unique-ai:unique-ai:sdk-licence-20261026.
enterprise_features_closed Are enterprise features in a separate closed edition? [unknown]: Current closed/open enterprise-edition boundaries were not established from an applicable component licence inventory. Public SDK/repository availability is not proof of the whole platform licence. Rechecked 2026-10-26 in Unique AI governance framework: contract access, Terms and Conditions Administration. No current customer MSA, quote, private schedule or gated contract was requested. The differently scoped 2023 sales-conversation ToS remains excluded.

Checked 2026-10-26; agent cross-checked 2026-10-26; check again by 2026-12-25. Sources: unique-ai:unique-ai:review2-governance, unique-ai:unique-ai:review2-terms-admin, unique-ai:unique-ai:reference-licence-20261026, unique-ai:unique-ai:unique-licence-v1-20261026, unique-ai:unique-ai:sdk-licence-20261026, unique-ai:unique-ai:sdk-package-20261026, unique-ai:unique-ai:review2-governance-20261026, unique-ai:unique-ai:review2-terms-admin-20261026.

VEN-06 Data export and exit

Judgment rule: Record admin export of chats, assistants, knowledge bases, prompts and audit logs, formats, export completeness, user self-export, and what survives contract end. Run an admin export on a trial. Exported chats alone do not make assistants and knowledge portable.

Unique AI

Agent cross-checked Support: partial

Authorised content/report export exists, but full assistant/knowledge portability is unverified.

Limitations

Reusable assistant/prompt and full knowledge/source exports, complete machine-readable state and current AI termination window remain unverified. No admin export fixture was run.

chat_export Can admins export all chats? [partial]: Detailed interaction export includes user prompts/answers over a chosen period with privacy settings. Completeness across files, agent state and all historical chats was not tested. Sources: unique-ai:unique-ai:analytics.
assistant_export Can assistants and prompts be exported in a reusable format? [unknown]: Can assistants and prompts be exported in a reusable format? Unknown. Reusable assistant/prompt and full knowledge/source exports, complete machine-readable state and current AI termination window remain unverified. No admin export fixture was run.
knowledge_export Can knowledge bases and their sources be exported? [unknown]: Can knowledge bases and their sources be exported? Unknown. Reusable assistant/prompt and full knowledge/source exports, complete machine-readable state and current AI termination window remain unverified. No admin export fixture was run.
export_format Are formats documented and machine-readable? [partial]: Model usage CSV and detailed analytics exports are documented. DPA promises return in existing format, not a universal reusable assistant/knowledge format. Sources: unique-ai:unique-ai:cost, unique-ai:unique-ai:analytics, unique-ai:unique-ai:dpa.
post_termination_window How long can data be retrieved after termination? [partial]: Baseline ToS 11.4 offers recorded-call download/export on written request within 30 days; that does not establish a chat/assistant/KB exit window. DPA return is within reasonable time. Sources: unique-ai:unique-ai:terms, unique-ai:unique-ai:dpa.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:analytics, unique-ai:unique-ai:cost, unique-ai:unique-ai:dpa, unique-ai:unique-ai:terms.

VEN-07 Support and service levels

Judgment rule: Record support channels, languages (German, French, Italian), hours and time zone, local or regional presence, documented response times, uptime SLA and credits, public status page and incident history. Popularity is not a reliability score.

Unique AI

Agent cross-checked Support: partial

Enterprise support and a public multitenant status page are available; response times are contracted.

Limitations

Guaranteed DE/FR/IT support, numeric hours/response times, dedicated uptime/credits and complete incident-history export were not established.

support_channels Which support channels are available on each plan? [supported]: Enterprise email and three-layer dedicated support escalation through customer team, CS and engineering. Hours/responses depend on SLA; extended-hours contract is extra. Sources: unique-ai:unique-ai:support.
support_languages Is support available in German, French or Italian? [unknown]: Is support available in German, French or Italian? Unknown. Guaranteed DE/FR/IT support, numeric hours/response times, dedicated uptime/credits and complete incident-history export were not established.
regional_presence Is there a DACH or Swiss entity, team or partner? [supported]: Unique AI AG lists Zürich, Unique Germany GmbH Berlin, plus UK/US/Singapore entities. Sources: unique-ai:unique-ai:contact.
uptime_sla Is there an uptime SLA with credits? [partial]: Support docs defer hours and response times to SLA contract; no public uptime/credit schedule was obtained. Baseline terms only promise commercially reasonable availability efforts. Sources: unique-ai:unique-ai:support, unique-ai:unique-ai:terms.
status_page Is there a public status page with history? [supported]: status.prod.unique.app publicly shows component/uptime history. Support docs scope it to multitenant accounts, not every dedicated environment. Sources: unique-ai:unique-ai:support, unique-ai:unique-ai:status.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:support, unique-ai:unique-ai:contact, unique-ai:unique-ai:terms, unique-ai:unique-ai:status.

VEN-08 Onboarding, training and adoption

Judgment rule: Record vendor-led onboarding, admin training, end-user training (live, e-learning, languages), use-case libraries, champion programmes, change-management materials and their cost. Distinguish included services from paid professional services and partner offerings.

Unique AI

Agent cross-checked Support: partial

The vendor markets forward-deployed delivery and structured training; inclusion and cost remain unpriced.

Limitations

Champion/adoption programme structure, training languages, included versus professional-service fees and onboarding hours are not published.

onboarding_service Is guided onboarding included? [partial]: Homepage markets embedded forward-deployed experts helping business/IT ship use cases; exact licence-included rollout scope is contracted. Sources: unique-ai:unique-ai:homepage.
user_training Is end-user training offered, in which languages? [partial]: Vendor homepage describes structured programmes teaching users to use/build/scale; current languages and course entitlement are not specified. Sources: unique-ai:unique-ai:homepage.
use_case_library Are role-based use cases and templates provided? [supported]: Public documentation has finance-role workflows, Chat/Conduct guidance, configurable Suggested Prompts and agent examples. Sources: unique-ai:unique-ai:platform, unique-ai:unique-ai:agent, unique-ai:unique-ai:conduct.
adoption_programme Is there a champion or adoption programme? [unknown]: Is there a champion or adoption programme? Unknown. Champion/adoption programme structure, training languages, included versus professional-service fees and onboarding hours are not published.
service_cost Which services cost extra? [unknown]: Which services cost extra? Unknown. Champion/adoption programme structure, training languages, included versus professional-service fees and onboarding hours are not published.

Checked 2026-10-21; agent cross-checked 2026-10-21; check again by 2026-12-20. Sources: unique-ai:unique-ai:platform, unique-ai:unique-ai:homepage, unique-ai:unique-ai:agent, unique-ai:unique-ai:conduct.

Sources

Unique LLM availability overview unique-ai:unique-ai:models
Lists contract-dependent provider/model availability, region SKUs, dated additions and retirements; CH Phoeniqs remains QA-only.
Per-client availability; Azure/LiteLLM/Phoeniqs matrices; Recent changes · documented · Source date 2026-10-12 · Retrieved 2026-10-21
Unique AI space configuration unique-ai:unique-ai:agent
Documents admin instructions, configured model/tools, knowledge binding, prompts and people/group Can chat/Can manage access.
Models; sources/tools; sharing; publishing · documented · Source date 2026-06-18 · Retrieved 2026-10-21
Unique AI Conduct unique-ai:unique-ai:conduct
Documents autonomous Python/Bash work, visible steps, approvals and stop, with per-chat files and restricted egress.
GA release 2026.40; execution UI; limitations · documented · Source date 2026-09-30 · Retrieved 2026-10-21
Unique release 2026.42 unique-ai:unique-ai:release42
Hosted release 14 October 2026 makes chat sharing and Context Memory GA, labels Scheduled Tasks experimental, changes protected-file ingestion and documents current model additions/migrations.
Release Overview; Chat Platform; RAG; Unique AI; Language Model · documented · Source date 2026-10-14 · Retrieved 2026-10-21
Custom LXM connection unique-ai:unique-ai:customllm
Documents operator-created external model endpoints, API/OAuth credentials and model groups.
Technical insight; Create and Connect; Security · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Code Execution unique-ai:unique-ai:code
Documents Azure Responses API Python execution, downloadable outputs, no outbound internet and model/file restrictions.
Compatibility; output; limitations; governance · documented · Source date 2026-07-30 · Retrieved 2026-10-21
Compliance Layer 3.0 unique-ai:unique-ai:compliance
Describes FINMA/bank-secrecy/CH-DPA Microsoft amendments and deployment-specific assurance limitations.
Legal framework; FSI amendments; deployment feature matrix · documented · Source date 2026-05-27 · Retrieved 2026-10-21
Custom embedding models unique-ai:unique-ai:embeddings
Documents custom embedding endpoints, one model per organisation and mandatory re-embedding after model/dimension changes.
Setup; configuration; re-embedding · documented · Source date 2026-05-07 · Retrieved 2026-10-21
On-premises security architecture unique-ai:unique-ai:onpremsecurity
Documents full customer control and possible operation without Unique access, with customer models/infrastructure.
Overview; custom setup · documented · Source date 2026-05-08 · Retrieved 2026-10-21
IT operator documentation overview unique-ai:unique-ai:operators
Describes dedicated Unique-managed, customer-cloud and full air-gapped on-premise modes.
Supported deployment models · documented · Source date not stated · Retrieved 2026-10-21
Code Execution unique-ai:unique-ai:review2-code
The page allows selected Knowledge Base files to be preloaded with additionalUploadedDocuments. Its compatibility list includes Azure and LiteLLM OpenAI Responses-capable IDs despite Azure-only introduction/limitations. Azure sandbox controls are not proof of the LiteLLM execution region.
Functionality; Limitations; All available models; Security & Governance; additionalUploadedDocuments · documented · Source date 2026-07-30 · Retrieved 2026-10-21
Unique LLM availability overview unique-ai:unique-ai:review2-models
The model catalogue separates provider routes, region/SKU choices and QA-only Swiss Phoeniqs. It is a technical availability inventory, not a complete legal subprocessor schedule.
Provider and region matrices; Phoeniqs Model Service; recent changes · documented · Source date 2026-10-12 · Retrieved 2026-10-21
Model usage and cost management unique-ai:unique-ai:cost
Documents off-by-default estimated usage dashboards and enforced daily company/group/user model-call limits.
Beta flags; daily limits; cost calculation; CSV · documented · Source date 2026-07-21 · Retrieved 2026-10-21
Upload and Chat limits unique-ai:unique-ai:files
Specifies 512 MB/file, 30 files/batch, default forced retrieval for selected uploads and configurable expiry.
Production limits; retrieval; expiry · documented · Source date 2026-06-28 · Retrieved 2026-10-21
Supported file types unique-ai:unique-ai:filetypes
Lists common Office/PDF/text formats and general ingestion limitations that differ from the newer chat-specific limits.
Types; exceptions; ingestion quality · documented · Source date 2026-05-09 · Retrieved 2026-10-21
Unique file ingestion administration unique-ai:unique-ai:file-ingestion-20261026
Base ingestion lacks scanned-PDF OCR. MDI supports image-based PDFs; Docling has partial support. MDI must be provisioned and enabled per folder or chat Space, and the guide distinguishes experimental image-content extraction and deprecated agentic ingestion.
Ingestion overview capability table; Enable MDI; Enable MDI for Upload in Chat · documented · Source date not stated · Retrieved 2026-10-26
Knowledge scope permissions unique-ai:unique-ai:kbpermissions
Describes query filtering by user/group scopes, but older SharePoint permission statements differ from the newer connector guide.
Scopes; source integration limits · documented · Source date 2026-05-07 · Retrieved 2026-10-21
SharePoint connector unique-ai:unique-ai:sharepoint
Documents periodic indexed sync, optional content-and-permissions mode, site selection and absent label-level detection.
Modes; permission sync; limitations · documented · Source date 2026-07-06 · Retrieved 2026-10-21
Confluence connector unique-ai:unique-ai:confluence
Documents Cloud/Data Center content sync without source-permission synchronisation.
Requirements; sync; limitations · documented · Source date 2026-06-28 · Retrieved 2026-10-21
Unique platform overview unique-ai:unique-ai:platform-overview-20261026
The platform architecture names Google Drive and Salesforce among knowledge sources, alongside SharePoint/OneDrive, Confluence and Teams/Outlook. It does not provide per-connector consent, rollout or price entitlements.
Connectors and Knowledge Sources diagram · documented · Source date not stated · Retrieved 2026-10-26
Unique release 2026.40 unique-ai:unique-ai:release40-20261026
GA HTML dashboards can list, enable and sign in to MCP connectors, with Salesforce and Yahoo Finance named examples. Contract/module enablement and personal connection state still apply.
Hosted September 30 2026; Chat Experience, GA HTML connector dashboards · documented · Source date 2026-09-30 · Retrieved 2026-10-26
Permission cache timing unique-ai:unique-ai:permissioncache
Describes usual subsecond internal permission propagation without a hard upper bound, and stale rules after a missed broadcast.
Expected timing; degraded scenarios · documented · Source date 2026-09-22 · Retrieved 2026-10-21
Single-tenant key management unique-ai:unique-ai:keys
Explains why client-held keys are unavailable across all Single Tenant data stores and directs full key control to self-hosting.
Big picture; BYOK/CYOK/HYOK; storage locations · documented · Source date 2026-05-27 · Retrieved 2026-10-21
Retention and deletion at Unique unique-ai:unique-ai:retention
Documents no training, Azure abuse-monitoring opt-out, post-termination retention and backup exceptions.
Hosted-only scope; Azure provider retention; deletion; encryption · documented · Source date 2026-05-04 · Retrieved 2026-10-21
Context Memory in Unique AI Chat unique-ai:unique-ai:memory
Documents private per-user cross-chat memory, edit/disable controls and an explicit warning that disabling does not delete the stored profile.
Private profile; edit and disable; web-search restriction · documented · Source date 2026-09-15 · Retrieved 2026-10-21
Unique AI Chat unique-ai:unique-ai:chat
Describes the user chat workspace and configurable tools.
Features and access · documented · Source date 2026-09-03 · Retrieved 2026-10-21
Access role concept unique-ai:unique-ai:roles
Defines identity, application, space and knowledge-scope permissions, including delegated management roles.
Four access layers · documented · Source date 2026-05-07 · Retrieved 2026-10-21
MCP tools for spaces unique-ai:unique-ai:mcp
Describes feature-flagged beta MCP clients with organisation and per-space connector/tool controls.
Beta warning; flags; selective tool enablement · documented · Source date 2026-05-21 · Retrieved 2026-10-21
MCP Hub security governance unique-ai:unique-ai:hub
Documents per-user token routing, approved origin restrictions and tool-call correlation; some DLP/elicitation pipeline measures are planned.
Authentication; governance; planned request pipeline · documented · Source date 2026-08-26 · Retrieved 2026-10-21
Deep Research tool unique-ai:unique-ai:deepresearch
Documents experimental research/report engine, with separate external OpenAI-engine data flows and limited current tools.
Experimental warning; engines; user flow · documented · Source date 2026-05-09 · Retrieved 2026-10-21
Web Search tool unique-ai:unique-ai:websearch
Documents multiple search engines, source citations and experimental query review/screening; conversation context may leave through queries.
Modes; engines; external-query warning · documented · Source date 2026-05-20 · Retrieved 2026-10-21
Python SDK documentation unique-ai:unique-ai:sdk
Documents SDK/public API access and customer-operated modules with event-driven webhook integration.
Public API; webhook verification; personal/app keys · documented · Source date 2026-05-08 · Retrieved 2026-10-21
Scheduled Tasks unique-ai:unique-ai:scheduled
Documents off-by-default recurring tasks running with user permissions; unattended approvals/questions can stop a run.
Beta warning; task settings; runs and limits · documented · Source date 2026-10-12 · Retrieved 2026-10-21
Scheduled Tasks unique-ai:unique-ai:review2-scheduled
Scheduled Chat/Conduct tasks offer Automatic or Manual approval. The newer release labels scheduling Experimental and feature-flagged; unattended automatic approval is not mandatory human confirmation.
At a glance; task Approval setting; unattended runs · documented · Source date 2026-10-12 · Retrieved 2026-10-21
Unique release 2026.42 unique-ai:unique-ai:review2-release42
Release 2026.42 makes Shared Chats GA, on by default per company with a per-Space toggle and preservation of explicit settings. Scheduled Tasks remain Experimental and feature-flagged.
Highlights; Chat Platform; Conduct scheduling flag · documented · Source date 2026-10-14 · Retrieved 2026-10-21
Knowledge Base MCP user guide unique-ai:unique-ai:mcpkb
Documents external-assistant search/read with user permissions, citations and optional KB scope restrictions.
Connecting; access; limitations · documented · Source date 2026-10-05 · Retrieved 2026-10-21
API model usage tracking unique-ai:unique-ai:api-usage
Documents public API/SDK model-call usage tracking; some other internal-service calls are outside scope and REST usage retrieval is planned.
Coverage; flags; limitations · documented · Source date not stated · Retrieved 2026-10-21
Translator space unique-ai:unique-ai:translation
Documents beta Office-file/text translation with glossary pre/postprocessing and potential layout disruption.
Beta; glossary; document types; limitations · documented · Source date 2026-05-06 · Retrieved 2026-10-21
Voice administration unique-ai:unique-ai:voice
Requires Azure AI Speech Services and describes potential public/backbone audio routing.
Dependencies; networking; limitations · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Unique data processing addendum unique-ai:unique-ai:dpa
Public DPA covers hosted chat/recordings, lists Microsoft, Swiss storage, return/deletion, confidentiality and ten-year audit-log retention.
Sections 2-3, 6-7; Appendices A-B; July 2025 footer · documented · Source date not stated · Retrieved 2026-10-21
Using Voice unique-ai:unique-ai:voice-user
Documents prompt dictation with EN/DE/FR/IT, immediate audio discard and dialect/recognition limitations.
Prompt dictation; privacy note; supported languages · documented · Source date not stated · Retrieved 2026-10-21
Single sign-on setup unique-ai:unique-ai:sso
Documents Zitadel SSO across deployment types, group claims and automatic SAML account creation/update.
Entra ID OIDC; SAML; managed-tenant setup · documented · Source date 2026-05-07 · Retrieved 2026-10-21
SCIM provisioning unique-ai:unique-ai:scim
Documents Unique SCIM endpoint/key for IdP user/group synchronisation.
Prerequisites and Entra setup · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Analytics interface unique-ai:unique-ai:analytics
Documents adoption reports, detailed chat/model/tool exports and randomised or stable pseudonymous identifiers.
Reports; anonymisation; detailed interactions · documented · Source date 2026-09-03 · Retrieved 2026-10-21
Audit logs unique-ai:unique-ai:audit
Defines privileged API event logging with self-service exclusions, deployment-specific retention and request-based export.
Scope; exclusions; default basic mode; retention and export · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Security Events Export unique-ai:unique-ai:security-export
Documents Single-Tenant-only Azure Activity/Defender Event Hub export with seven-day replay, client-region selection and consumption cost.
Availability; delivered events; retention; cost · documented · Source date not stated · Retrieved 2026-10-21
Audit logs unique-ai:unique-ai:review2-audit
API audit-event guide specifies basic/verbose modes and MT 90-day/ST five-year operational retention. It does not reconcile those event classes to DPA audit logs or a customer-specific contractual override.
About audit logs; modes; Single and Multi Tenant Clients · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Unique data processing addendum unique-ai:unique-ai:review2-dpa
The public July 2025 DPA specifies ten-year audit logs and CH storage, while Appendix A lists Microsoft and a client-chosen location. No signed current provider schedule, retention override or contract-precedence mapping was inspected.
Sections 3.15 and 6.1; Appendix A; July 2025 footer · documented · Source date not stated · Retrieved 2026-10-21
Retention and deletion at Unique unique-ai:unique-ai:review2-retention
Hosted-only policy separates ordinary post-termination content retention, recoverable files and backups and permits contract customisation. Its log/key statements are not reconciled to audit-event/DPA/key-management scopes.
Purpose; Data retention; Deletion after termination; Backups; Encryption in Place · documented · Source date 2026-05-04 · Retrieved 2026-10-21
Data leakage prevention integration unique-ai:unique-ai:dlp
Describes customer-operated HTTPS/DLP proxy and post-chat analytics scanning rather than a default native PII-redaction feature.
Proxy; analytics API; compatibility · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Single-tenant security architecture unique-ai:unique-ai:single
Documents separate Azure landing zone, network restrictions, retention and Unique-managed infrastructure keys.
Azure resources; SSO; encryption; monitoring · documented · Source date 2026-05-28 · Retrieved 2026-10-21
Terms and conditions administration unique-ai:unique-ai:policy
Documents first-login acceptance, custom terms under an MSA and version/agreement export through GraphQL.
First-login acceptance; MSA condition; API export · documented · Source date 2026-05-07 · Retrieved 2026-10-21
Single-tenant key management unique-ai:unique-ai:review2-keys
May 2026 key guide distinguishes imported customer-generated keys held by Unique from customer-exclusive key control via self-hosting. Its MT/ST CMK statement differs from the hosted retention page; actual MT configuration is unverified.
Infrastructure-only scope; PMK; BYOK/CYOK/HYOK; Summary · documented · Source date 2026-05-27 · Retrieved 2026-10-21
Single-tenant security architecture unique-ai:unique-ai:review2-single
Single-tenant guide documents separate landing zone, WAF/IP controls and Unique-managed encryption. Its 2048-bit RSA-HSM description differs from the hosted retention page; this does not establish an exact current key design.
Overview; Encryption; Data classification · documented · Source date 2026-05-28 · Retrieved 2026-10-21
Compliance Layer 3.0 unique-ai:unique-ai:review2-compliance
The deployment matrix documents CH North multitenant hosting and selectable single/client-hosted regions, with deployment-specific controls. These are options, not proof of a particular customer contract or current key configuration.
Digital Sovereignty; Feature overview matrix · documented · Source date 2026-05-27 · Retrieved 2026-10-21
Unique installation models unique-ai:unique-ai:install
Enumerates multi-tenant, single-tenant, customer-managed and on-premise deployments.
Overview · documented · Source date 2025-08-06 · Retrieved 2026-10-21
Unique terms of service unique-ai:unique-ai:terms
Describes a sales-conversation service with Order Form subscription fees, nonrenewal notice, recording export and proprietary rights; applicability to the current AI MSA needs confirmation.
Sections 1-2, 4-6, 11, 17-18 · documented · Source date not stated · Retrieved 2026-10-21
Chat data redaction and retention unique-ai:unique-ai:redaction
Documents scheduled clearing of chat content and workspace records, with separate upload/KB retention rules.
What is cleared; exclusions; administrator settings · documented · Source date 2026-10-18 · Retrieved 2026-10-21
Unique AI contact unique-ai:unique-ai:contact
Lists Unique AI AG in Zürich, Unique Germany GmbH in Berlin and other regional entities.
Office addresses · documented · Source date not stated · Retrieved 2026-10-21
Unique AI governance framework: contract access unique-ai:unique-ai:review2-governance
The current governance framework says legal contracts and end-user T&Cs are available on request; the public page does not expose platform subscription or component licence clauses.
Individual standards; Legal contracts; End-User Terms and Conditions · documented · Source date 2026-04-12 · Retrieved 2026-10-21
Unique AI trust centre unique-ai:unique-ai:trust
Describes CH Azure storage separately from globally routed inference and claims ISO/SOC assurances, with differing SOC types in the FAQ.
Enterprise security; compliance standards; FAQ · vendor-claimed · Source date not stated · Retrieved 2026-10-21
Unique ISO/IEC 42001 certificate unique-ai:unique-ai:iso42001
TÜV SÜD certificate covers development/deployment SaaS for financial services using pretrained foundation models or client-provided models, valid to 6 March 2028.
One page; registration 99 420 00001; scope and dates · certificate-verified · Source date 2025-03-07 · Retrieved 2026-10-21
Issuer: TÜV SÜD South Asia Private Limited. Scope: Unique AG, Stockerstr.34, Zurich. Development and deployment of SaaS for financial-services clients based on pretrained AI foundation models using Azure OpenAI or client-provided models, including client on-premise models.. Valid 2025-03-07 to 2028-03-06
ISO Certifications and current downloads unique-ai:unique-ai:review2-iso-docs
The ISO documentation links 2026 English/German ATTESTA ISO27001 and ISO9001 renewals, plus the 2025 ISO42001 certificate; ISO27001 SoA requires an NDA.
Certificate ISO27001; ISO9001; ISO42001; Statement of Applicability · documented · Source date 2026-08-05 · Retrieved 2026-10-21
Unique AG ISO27001 ATTESTA 2026 renewal unique-ai:unique-ai:review2-iso27001
Inspected ATTESTA ISO27001:2022 renewal for Unique AG, registration 223-06-018.I, issued 10 June 2026 after 4 June surveillance, valid until 31 May 2029. The certificate-linked public register lists the same number and expiry.
One-page 2026 English certificate; scope, registration, surveillance, printed expiry · certificate-verified · Source date 2026-06-10 · Retrieved 2026-10-21
Issuer: ATTESTA Schweizer Zertifizierungsgesellschaft AG. Scope: Unique AG, Stockerstrasse 34, 8002 Zürich: development, operation, trading and distribution of software and provision of IT/information-technology services; ISO27001:2022 information security management. Valid from an unstated date to 2029-05-31
Unique AG ISO9001 ATTESTA 2026 renewal unique-ai:unique-ai:review2-iso9001
Inspected ATTESTA ISO9001:2015 renewal for Unique AG, registration 223-06-017.Q, issued 10 June 2026 after 4 June surveillance, valid until 31 May 2029. The certificate-linked public register lists the same number and expiry.
One-page 2026 English certificate; scope, registration, surveillance, printed expiry · certificate-verified · Source date 2026-06-10 · Retrieved 2026-10-21
Issuer: ATTESTA Schweizer Zertifizierungsgesellschaft AG. Scope: Unique AG, Stockerstrasse 34, 8002 Zürich: development, operation, trading and distribution of software and provision of IT/information-technology services; ISO9001:2015 quality management. Valid from an unstated date to 2029-05-31
Certificate-linked public register: Unique AG unique-ai:unique-ai:review2-iso-register
Public certificate-linked register lookup for Unique AG lists registrations 223-06-018.I and 223-06-017.Q, Zurich, software/IT scope and expiry 31 May 2029. Standard/status text fields are blank; certificate PDFs supply the standards.
Unique AG result rows; registration, scope, expiry · documented · Source date not stated · Retrieved 2026-10-21
SOC2 Type2: chronology and report access unique-ai:unique-ai:review2-soc
The dedicated SOC page explains Type1 in December 2023 followed by Type2 in January 2025; audit report and certificate are available only on request after NDA. Auditor, service scope and actual reporting period were not publicly inspected.
SOC2 Type2 timeline; Certificates/report access · vendor-claimed · Source date 2026-04-12 · Retrieved 2026-10-21
Terms and Conditions Administration unique-ai:unique-ai:review2-terms-admin
Current platform T&C administration refers to a customer MSA and gated admin export; it does not publish the vendor platform agreement or determine subscription/IP/branding clauses.
Overview; MSA exceptions; Modifying/Exporting Terms and Conditions · documented · Source date 2026-05-08 · Retrieved 2026-10-21
Unique AI platform unique-ai:unique-ai:platform
Markets a modular finance platform with governed connectors, frontend, financial agents and APIs.
Frontend, financial agents, MCP Hub; deployment section · vendor-claimed · Source date not stated · Retrieved 2026-10-21
ECB euro foreign exchange reference rates, 20 October 2026 unique-ai:unique-ai:ecb-fx
ECB records EUR 1 = USD 1.1215 and CHF 0.9237 on 20 October 2026; derived USD to EUR is 1/1.1215 and CHF to EUR is 1/0.9237.
Cube time="2026-10-20": USD and CHF · documented · Source date 2026-10-20 · Retrieved 2026-10-21
Unique AI governance framework: contract access unique-ai:unique-ai:review2-governance-20261026
The current governance framework says legal contracts and end-user T&Cs are available on request; the public page does not expose platform subscription or component licence clauses.
Individual standards; Legal contracts; End-User Terms and Conditions · documented · Source date 2026-04-12 · Retrieved 2026-10-26
Unique public reference repository licence unique-ai:unique-ai:reference-licence-20261026
The public repository defaults to Unique License v1 unless an explicit licence says otherwise; commercial use, distribution and modifications require prior approval. The default is restricted source-available, not an OSI open-source licence.
LICENSE, repository-wide disclaimer and exceptions · documented · Source date not stated · Retrieved 2026-10-26
Unique License v1 unique-ai:unique-ai:unique-licence-v1-20261026
The published full licence permits personal noncommercial use and requires prior written approval for commercial use, distribution and modification, with separate component-licence exceptions. Swiss law and Zurich venue are specified.
Grant, Restrictions, Open Source Components, Governing Law · documented · Source date 2024-09-24 · Retrieved 2026-10-26
Unique Python SDK MIT licence unique-ai:unique-ai:sdk-licence-20261026
The Python SDK directory has an explicit MIT licence permitting commercial use and modification subject to notice requirements; this exception does not license the hosted platform.
unique_sdk/LICENSE, explicit component exception · documented · Source date not stated · Retrieved 2026-10-26
Unique SDK package metadata unique-ai:unique-ai:sdk-package-20261026
At the pinned October 23 repository commit, unique_sdk package metadata declares MIT, consistent with its component licence. Repository/package version is not treated as the hosted release date.
project name/version and licence · documented · Source date not stated · Retrieved 2026-10-26
Terms and Conditions Administration unique-ai:unique-ai:review2-terms-admin-20261026
Current platform T&C administration refers to a customer MSA and gated admin export; it does not publish the vendor platform agreement or determine subscription/IP/branding clauses.
Overview; MSA exceptions; Modifying/Exporting Terms and Conditions · documented · Source date 2026-05-08 · Retrieved 2026-10-26
Enterprise support unique-ai:unique-ai:support
Provides public multitenant status link, enterprise email and a contract-specific dedicated support model.
Contact; dedicated support model; hours · documented · Source date not stated · Retrieved 2026-10-21
Unique public system status unique-ai:unique-ai:status
Public status page displays service/API/collector components and historical uptime panels; dedicated environments are outside the documented status scope.
Component and 90-day uptime panels · documented · Source date not stated · Retrieved 2026-10-21
Unique AI finance platform unique-ai:unique-ai:homepage
Markets embedded experts and structured training programmes for finance users and IT.
Forward-Deployed Engineers and training section · vendor-claimed · Source date not stated · Retrieved 2026-10-21