Timeline

Reuters reveals leaked Meta document permitted AI chatbots romantic conversations with children

The 200-page standards document, signed off by Meta's legal, policy and chief ethicist, also permitted racist arguments framed as factual and was later called an internal error.

  • Security & misuse
  • Culture & impact
  • Major

Reuters reported that it had reviewed an internal Meta document, “GenAI: Content Risk Standards,” running to roughly 200 pages, which set out example conversations that Meta’s AI chatbots were permitted to have. Among the provisions: chatbot personas were allowed to “engage a child in conversations that are romantic or sensual,” though not to describe sexual acts to minors during roleplay. Reuters cited one approved sample response to a high-schooler asking “what are we going to do tonight, my love?” that included romantic language about physical closeness. The document also permitted chatbots to generate demeaning statements about people based on protected characteristics — including comparisons of intelligence between racial groups — if the statements were framed as factual argument rather than harassment. Meta confirmed the document was authentic and said it had been approved by its legal, public policy and engineering staff, and by the company’s chief ethicist.

Meta spokesperson Andy Stone disputed that the guidelines sanctioned “provocative behavior with children,” saying the passages permitting romantic chat with minors were “erroneous and inconsistent” notes that had already been removed and did not reflect company policy, and that Meta now prohibited flirtatious or romantic exchanges with minors outright. A child-safety advocate quoted in the reporting rejected the reassurance and called on Meta to publish its current guidelines so parents could verify the change independently.

The report prompted Senator Josh Hawley, chair of a Senate Judiciary subcommittee, to open an investigation into whether Meta’s generative-AI products had enabled harm to children or whether the company had misled regulators about its safeguards, demanding Zuckerberg preserve related records. It also fed into a broader reckoning: the following month the FTC opened a formal inquiry into AI companion chatbots’ effects on children covering Meta, Alphabet, OpenAI, xAI, Snap and Character Technologies.