What it is
Yone AI is a routing layer. It accepts a request, selects or honours a specified AI model, forwards the request to that model's provider, and returns the response. It does not host or operate any AI model itself.
What it does not do
- Train, fine-tune, or otherwise modify any AI model
- Author, inject, or alter your prompts
- Evaluate, rank, or filter model outputs on quality
- Make or influence any automated decision producing legal or similarly significant effects on a data subject
- Retrieve data from external or public sources — it retrieves data from within your own organisation's content only, via tool-calling, unless you separately use the named web-search feature where available
This is the scope boundary of the system: output review, and any decision made using AI-generated content, is the customer's responsibility (see "Human oversight" below).
Providers
| Provider | Legal entity | Path |
|---|---|---|
| Anthropic | Anthropic, PBC, 548 Market Street, PMB 90375, San Francisco, California 94104, United States | Direct API |
| Microsoft | Microsoft Limited, Microsoft Campus, Thames Valley Park, Reading, Berkshire RG6 1WG, United Kingdom | Azure AI Foundry |
Full model-by-model detail (which model runs on which provider and deployment type) is maintained in the DPA Annex III Part D — AI Model Routing Table, which is the source of truth this card refers to rather than restates.
Tiers
- Global (shared) — the default. Shared platform infrastructure and capacity across all Customers.
- Dedicated key — a client-specific model configuration and key. Available on request, administrator-provisioned, not self-service.
- Dedicated deployment — an isolated environment with guaranteed capacity. Available on request, subject to a separate agreement, not a standing self-service feature.
- BYOK — your own provider account and key.
See Terms of Service Clause 8.4.9 for the customer-facing description of these tiers.
Processing
Prompts are held in memory in transit to the provider and back; see the Terms of Service Clause 17.10 for what is separately logged (a short excerpt of the initial prompt, for administrative and abuse-monitoring purposes — not the full prompt, and never the AI's response). Processing location is set out in the Routing Table referenced above. Failover: off for interactive AI Features; a platform-default model may be substituted for background AI functions only (KB generation, agent-spec generation, document analysis) if a client's own model configuration is invalid — see Terms of Service Clause 8.4.8.
Human oversight
None by us. AI Features are decision-support tools; all output must be reviewed and validated by a qualified human before being acted upon. Output review, and any decision made using AI-generated content, sits outside this system's boundary and is the Customer's responsibility (Terms of Service Clause 17.5).
Known limitations
- Output accuracy is determined by the selected model, not by this system.
- Provider availability and behaviour are outside our control.
- Model versions may change on the provider's own schedule.
Out of scope uses
AI Features are provided within the business-process, enterprise-architecture, and requirements-management scope of the Service (Terms of Service Clause 17.9). You must not use AI Features to generate content that is unlawful, harmful, discriminatory, deceptive, or that violates third-party rights, or for automated decision-making with legal or similarly significant effects, or for regulated professional advice.
Change policy
Model additions, removals, and default-routing changes: governed by Terms of Service Clause 8.4.6 — at least 30 days' notice for changes that increase Token consumption or withdraw a model; reductions and additions may take effect immediately.