How to turn off Meta AI on Facebook – What you can and can't control (proton.me)

Meta has rolled its Meta AI assistant deeply into Facebook, Instagram, Messenger and WhatsApp — answering questions, summarizing chats, suggesting content and generating images — but there is no global “off” switch. Meta can access and process your AI chats, interacting posts and some public content to train its models, and human reviewers may see flagged material. In June 2025 some AI prompts were exposed on a public Discover feed, and Meta plans to start using chatbot conversations to target ads on Facebook and Instagram from December 16, 2025. Only users in jurisdictions with stronger privacy laws (EU, UK, Switzerland, Brazil, Japan, South Korea) currently have formal opt-out rights; most users worldwide, including in the US, do not. Practically, deleting chats or avoiding the Ask Meta AI buttons doesn’t guarantee your data won’t be used: opt-out requests via Meta’s Privacy Center apply only to future processing, won’t erase data already used for training, and won’t stop others from invoking @MetaAI in group chats that include your messages. The only surefire way to stop future interactions is to leave Meta’s ecosystem, but even that won’t remove previously ingested data. For users concerned about privacy, recommended mitigations are to avoid interacting with Meta AI, delete chats, submit region-appropriate opt-out requests, audit sharing settings, and consider privacy-first AI alternatives that explicitly do not log or reuse conversations.