Skip to content

Approved AI Provider List

This is the single source of truth for which AI providers Feoda employees may use, and for which categories of data. Any provider not on this list is not approved for Feoda work.

Status: Draft skeleton — pending evidence collection and ratification under AP-03. Do not rely on this list for production decisions until status is Approved.

Eligibility Tiers

Providers are evaluated against the data classification model in ai-strategy.md Section 8.

Tier Data It May Handle Minimum Provider Requirements
T1 — Public-only Public-tier data only Reasonable terms; no Feoda data of consequence
T2 — Internal Public + Internal Approved enterprise/business tier; SSO; named-individual accounts
T3 — Confidential Public + Internal + Confidential T2 + signed DPA; restricted to specific named workflows; full audit logging
T4 — Restricted Public + Internal + Confidential + Restricted T3 + region-appropriate data residency; legal review per workflow
Prohibited (none — never use this provider for Feoda work)

Approved Providers

Provider Product / Tier Approved Tier DPA Status SSO Status Audit Log Owner Contract End Next Review
(to be populated under AP-03)

Provisional Candidates (under evaluation)

Providers identified for evaluation but not yet approved. They may not be used for any tier above what their evaluation status permits.

Provider Product / Tier Target Tier Evaluation Status Notes
Anthropic Claude for Work / Enterprise T3 (target) To evaluate Used in current platform Phase 2 (api/chat.js)
Groq Cloud (free tier) T1 only To evaluate Free tier — Public data only; not for Internal+
OpenAI ChatGPT Enterprise / Team T3 (target) To evaluate
GitHub Copilot Business / Enterprise T3 (target) To evaluate

Prohibited Tools

Examples (illustrative, not exhaustive). Specific named prohibitions are added by the Head of Technology when identified.

  • Free / consumer tiers of any AI provider for any data above Public tier
  • Personal accounts (paid or free) on any AI provider for any Feoda work
  • Browser extensions or unverified third-party tools that route Feoda content to AI providers without explicit approval
  • Any provider that retains the right to train on customer inputs in its applicable terms

Process

  • Adding a provider: Submit evaluation under AP-03. Evidence required for the target tier (DPA, SSO, audit, residency as applicable). Final approval by Head of Technology.
  • Removing a provider: Documented reason; migration plan for affected workflows; communication; access disabled.
  • Exception use: No exceptions to this list. If a workflow needs a provider not on the list, the provider must be evaluated and added — not bypassed.
  • Review cadence: Quarterly. Each entry's Next Review column is updated on review.

Change Log (this document)

Date Change Author
2026-04-22 Initial skeleton created. Pending AP-03 evidence collection and ratification. Head of Technology