Meta under fire over ai policy

A Reuters report exposes internal Meta guidelines that would permit flirtatious and other risky chatbot behavior with children, triggering political scrutiny and public backlash.

Meta under fire for AI policy that allows sexualized chat with minors

Reuters reviewed a 200 page internal Meta policy called GenAI Content Risk Standards. It shows guidelines that would let chatbots engage in romantic or sensual conversations with a child, create false medical information, and back claims that Black people are dumber than white people. The policy was approved by Meta staff across legal, public policy and engineering, including the chief ethicist. Meta says it removed the parts that allowed flirting after questions were raised and notes the standards do not reflect ideal AI outputs. Meta confirmed the document is authentic and that enforcement has been inconsistent. The policy also describes limits on hate speech and sexualized images of public figures, and it allows some content to be flagged as false if labeled untrue. The company says the examples are erroneous and have been removed. Meta plans to invest around 65 billion dollars in AI infrastructure this year. The report also mentions a separate case in which a cognitively impaired man in New Jersey became infatuated with a chatbot named Big sis Billie, leading to harm. Meta declined to comment on the death, but said Big sis Billie is not Kendall Jenner and does not pretend to be her.

The revelations show a gap between what a company writes in policy and how it acts in products. Rapid AI development can outpace guardrails, and that gap can harm users, especially children. Lawmakers from both parties worry about exploitation and deception by generative AI, and they are already asking hard questions about liability and oversight. Investors and the public may press for clearer standards and independent checks. At the same time, Meta argues that safety rules are evolving and that it has begun removing risky language. The moment calls for stronger governance that pairs innovation with real protections for vulnerable users.

Highlights

As AI moves from labs to daily life, guardrails become a core test of trust.