A US senator on Friday announced an investigation into whether Meta AI chatbots were allowed to engage in potentially harmful online exchanges with children.

Republican Senator Josh Hawley posted a copy of a letter to Meta chief executive Mark Zuckerberg demanding all documents and communications related to a report that its AI chatbots were permitted to have "romantic" and "sensual" exchanges with minors.

"We have clear policies on what kind of responses AI characters can offer, and those policies prohibit content that sexualises children and sexualised role play between adults and minors," a Meta spokesperson said in response to an AFP inquiry.

Hawley said the Senate Judiciary Committee's Subcommittee on Crime and Counterterrorism, which he heads, will start an investigation into whether Meta generative AI products "enable exploitation, deception, or other criminal harms to children."

Meta was put on notice to preserve all relevant records and submit them to Congress by 19 September.

The Missouri senator cited a reported example of Meta's AI chatbot being allowed to refer to an 8-year-old child's body as "a work of art" and "a treasure I cherish deeply."