U.S. Senator Josh Hawley has launched an investigation into Meta Platforms in light of a leaked internal document that purportedly revealed the company's AI systems were allowed to engage in "sensual" or "romantic" conversations with children. The document, which was reportedly called "GenAI: Content Risk Standards" and was obtained by Reuters, included, among other things, scenarios whereby Meta's generative AI chatbots could create follow inappropriate responses, including describing the body of eight-year-old in sexualized ways. 

Hawley is a Missouri Republican and called the news "reprehensible and outrageous," and wrote on X that it appeared Meta was willing to program its chatbots to have explicit conversations with minors. Hawley stated that he will obtain this internal document and the details of related products, and added, "parents deserve the truth" and that children needed to be protected from technology that can be exploitative.

Meta stated from the outset that these claims were "erroneous and inconsistent" with its policies.  A Meta spokesperson explained that Meta's public guidelines prohibit sexualized interactions between children and adults and sexualized role play with minors, and that the notes leaked were from internal teams working through hypothetical situations, not actual policy. The spokesman indicated that the problematic material was "removed" at this point. 

The document also apparently indicated that according to Meta's guidelines, its chatbots could provide inaccurate medical information, share provocative conversations about sensitive issues such as race or sex, and could even share inaccurate information about celebrities, as long as they include a disclaimer. 

The leaks demonstrate the heightened scrutiny being aimed at how Big Tech will manage AI safety, especially with regard to children using Meta owned products such as Facebook, Instagram, and WhatsApp.