Opaque 'safety classifiers' let model providers retain your chats indefinitely

niloofar_mire · x · 2026-09-09

Part of a thread on data retention: providers' 'zero data retention' promises come with carve-outs — if an opaque, poorly understood 'safety classifier' flags your chat as a safety concern, the provider can keep it for as long as it wants.

Related event: Vague safety classifiers and court orders quietly extend AI data retention(2 posts)→

Original post →

More from Safety

Safety channel →