Provider Trust Tracker

A provider's own policy statement yields "claimed" at most. Only independently validatable evidence — a checkable attestation, a reproducible technical probe, on-chain data — yields "verified". Absence-claims ("zero data retention", "we don't log prompts") can never be verified from a policy statement alone; they stay "claimed" until there is cryptographic or attestation evidence to check.

In practice: a provider’s retention or training policy page can move a field to claimed at most. Moving a field to verified requires something we can independently check — a validatable attestation, a reproducible technical probe, or on-chain data. A row that says only “under review” means we have not yet completed that check, not that the provider has a bad policy. Full verification-grade definitions are on methodology.

Seeded from the roster of major closed and open-model-host providers plus DeAI’s Tier 1 decentralized/private inference providers (last updated 2026-08-24). Morpheus is included and graded on the identical rubric as every other row — as the publication’s backer it gets no benefit of the doubt; a favorable, unverified Morpheus claim would discredit the tracker. Feeds the programmatic “does [Provider] train on your data?” pages as those publish.