Free, Pro & Max: No BAA · Trains on your data by default unless you opt out · Outside SOC 2 and ISO scope
Team, Enterprise & API: BAA available but must be activated · Does not train by default · Six API features excluded from BAA coverage
| Claude Free, Pro & Max | Claude Team, Enterprise & API | |
|---|---|---|
| HIPAA BAA | NoBAA is offered only for commercial products; Anthropic explicitly excludes Claude Free, Pro and Max | Yesbut must be activated - Enterprise requires the Primary Owner to enable HIPAA compliance in a HIPAA-ready organization; API requires a signed BAA plus Anthropic enabling it. Batch API, Files API, Skills API, Code Execution, Computer Use and Web Fetch are excluded from BAA coverage |
| Trains on your data | Yesby default - inputs and outputs are used for training unless the user opts out in account settings; safety-flagged and user-reported content is used regardless of opt-out | Noby default - "Anthropic may not train models on Customer Content from Services"; training only if the customer opts in via feedback or explicit sharing |
| Survives deletion | Yesdeleted chats clear history immediately and back-end within 30 days, but training opt-in retains de-identified chats up to 5 years; policy-violation content up to 2 years and safety scores up to 7 years | Deleted conversations clear history immediately and back-end within 30 days; API inputs and outputs auto-delete within 30 days. Policy-violation content up to 2 years, safety scores up to 7 years. Enterprise custom retention controls are documented separately and not verified in this pass |
| Deployer / admin log access | Nodocumented deployer access - consumer accounts are individual, not organization-managed; no admin log, audit or eDiscovery path is published for these tiers | Not publicly stated in the retention documentationno admin export, audit log or compliance API is specified there |
| Third-party attestation | Out of scope - SOC 2 Type I & II, ISO 27001:2022 and ISO/IEC 42001:2023 are scoped to commercial products, explicitly distinguished from Free, Pro and Max | YesSOC 2 Type I & II, ISO 27001:2022 and ISO/IEC 42001:2023, scoped to commercial products (Claude for Work and the Anthropic API); reports via the Trust Portal |
Claude Team and Enterprise, which do not train by default and can carry a BAA; paying does not change tier class - Max at $100/month is still consumer and has less compliance coverage than a $20 Team seat
Claude Pro and Max, which are consumer tiers that train by default and carry no BAA; also the assumption that an executed BAA covers all API features, when six are excluded
Vendor terms change without notice. This page reflects what Claude published as of September 1, 2026. Re-verified quarterly.
If you need a scored Go / Conditional Go / No-Go verdict for a vendor decision — including a vendor not listed here — that's an AI Vendor Risk Assessment: three binary industry gates, nine scored criteria, four provenance fields on every finding. $997, delivered in 48 hours.
See how AVRA works →Want all 13 tools in one file?
Covers 22 tiers across 13 tools — same sourcing, side by side, printable for a vendor file.