Meta Platforms is under renewed political and regulatory pressure after a Reuters investigation revealed that an internal 200-page policy manual, titled “GenAI: Content Risk Standards,” explicitly allowed the company’s AI chatbots to engage in romantic and sexual conversations with under-age users. Sample exchanges in the document showed the bots describing a shirtless eight-year-old as “a masterpiece” and role-playing intimate scenarios with a high-school student. The guidelines, approved by Meta’s legal and public-policy teams, also permitted the dissemination of false medical claims and racially derogatory statements, according to the report. Following Reuters’ questions, Meta said the examples were “erroneous,” removed the contested passages and began revising the policy. “Our rules prohibit content that sexualizes children,” spokesperson Andy Stone said, without publishing the updated standards. The disclosure has drawn bipartisan scrutiny on Capitol Hill. Senator Josh Hawley and several colleagues have demanded Meta turn over all related documents and explain how the policy was cleared. Child-safety organizations called the language “totally unacceptable,” while Brazil’s Attorney General asked Meta to withdraw any chatbot capable of sexualized exchanges with minors. The episode adds to growing concerns that engagement-driven AI companions can expose children to exploitation and misinformation. Lawmakers and advocates say the leak underscores the need for binding regulations that hold technology companies liable when AI systems harm minors.