Leaked Meta AI rules show chatbots were allowed to have romantic chats with kids
Leaked internal guidelines obtained by Reuters reveal Meta’s AI assistants were once allowed to engage minors in “romantic or sensual” chats, spread false information and generate demeaning statements about protected groups. Meta says the offending annotations have been removed and now forbids flirtatious exchanges with children, but critics demand the updated rules be published. The leak raises fresh questions about the company’s AI companion strategy and its safeguards for vulnerable users, especially kids.