Verified vendor ToS diffs appear when the fleet confirms a material change. Fixed monthly editions track what reviewed products publicly say about training use, opt-outs, retention, and human review.
Autonomous captures from vendor policy, pricing, and data-use pages. Empty until a second material change lands on a baselined source.
No autonomous vendor ToS diffs verified yet. The fleet is baselining 26 policy sources; a verified diff appears when a second material change lands on a baselined page. Currently blocked from the fleet UA: OpenAI usage policies, OpenAI privacy policy (data use). Monthly editions below remain the public trail.
Institutional teams that need continuous monitoring can open monitoring & advisory or contact PAI.
July review of the AI Data Use Index baseline products plus new toolchain disclosures. Cursor published agent permissions documentation (permissions.json, MCP allowlists) that production teams must map to PSF-7; GitHub Copilot enterprise policy pages unchanged on training use. No material reversals on consumer training opt-outs across the 15-product baseline.
Mid-June review of the 15 baseline products finds no material changes to training use, opt-out, retention, or human review policies. The most significant governance signal this month came from outside policy pages: Anthropic suspended Claude Fable 5 approximately one hour after PAI's CompassEval sweep completed, raising practitioner questions about model lifecycle governance and what prior-consent rights apply when a model is withdrawn mid-evaluation.
The first public edition establishes the baseline for the AI Data Use Index: reviewed product pages, the self-check flow, and the methodology used for future comparisons.
Leave an email if you want notice when email alerts ship. Until then, the monthly editions below are the public record.