Frontier labs' ToS loopholes: a single thumbs-up can strip your chats of protection

niloofar_mire · x · 2026-09-10

Niloofar Mire dissects misleading ToS clauses at frontier model providers: data collection is opt-out by default (opt-in is the default), and 'explicit feedback'—even a single thumbs-up/down or picking the better response—can void claimed chat protections, making ZDR assumptions false even in enterprise settings. Quoted, MimeeXu argues AI shouldn't replicate ad-tech dark patterns and privacy should be a given, not individual responsibility.

Related event: Frontier AI Vendors' Privacy Terms Hide Data-Collection Traps(3 posts)→

Original post →

More from Companies & People

Companies & People channel →