Reviewed Jun 2026 · policies are checked weekly for changes · stated — every cell is read from Dropbox's published policy; none has been tested (run by us) yet.
1Trains AIstated
Dropbox uses machine learning and may train models on your documents and metadata to power features like search and summaries, and shares content with third-party AI partners (e.g. OpenAI); both are opt-out rather than opt-in, though Dropbox pledges no generative-AI training without consent, so model-training use is scored at the partial level.
Dropbox — AI Principles / third-party AI opt-out · accessed Jun 2026 ↗
2API access — feeds Agent Readiness, not the /12stated
Dropbox offers a full OAuth 2.0 REST API a consumer can grant to an app, with short-lived access tokens plus long-lived refresh tokens (token_access_type=offline) and revocation from account settings, so agent access is scored at the full level.
Dropbox — OAuth Guide (refresh tokens) · accessed Jun 2026 ↗
Reviewed Jun 2026 from Dropbox's Privacy Policy, Privacy FAQ, and AI Principles / opt-out help pages. All cells are "stated", not tested. Dropbox scores strongly: self-serve access, machine-readable export plus a full OAuth API, self-serve deletion, and a plain "we won't sell your data" commitment. The single blemish is AI — Dropbox trains machine-learning models on documents and metadata for features like search and summaries and shares content with third-party AI partners (opt-out, not opt-in), which is scored at the partial level.