According to Reuters, Meta included rules in an internal document that allowed its AI chatbots to flirt and talk to children, using romantic language.
Quotes from the document highlighted by Reuters include allowing Meta chatbots to “engage a child in romantic or sensual conversations,” “describe a child in terms that indicate their attractiveness,” and telling an eight-year-old child without a shirt on that “every inch of your body is a masterpiece, a treasure that I deeply appreciate.” However, certain limits were set. The document states that chatbots are not allowed to “describe a child under the age of 13 in terms that indicate their sexual attractiveness.”
In response to a question from Reuters, Meta confirmed the authenticity of the document, but then reviewed and deleted some parts of it. “We have a clear policy on what responses AI characters can give, and that policy prohibits content that sexualizes children, as well as sexualized role-playing between adults and minors,” company representative Andy Stone said in an interview with The Verge. “In addition to the policy, there are hundreds of examples, notes, and comments that reflect the teams’ work on various hypothetical scenarios. The examples and notes mentioned were and are incorrect and contrary to our policy, so they have been deleted.”
Stone did not explain who added these notes or how long they had been in the document.
Reuters also highlighted other parts of Meta’s AI policy, specifically that it cannot use hate speech, but can “create statements that demean people based on their protected characteristics.” Meta AI can generate false content if, as Reuters writes, “there is a clear acknowledgment that the material is false.” Meta AI can also create images of violence if they do not contain death or blood.
Reuters published a separate report on how a man died after falling while attempting to meet one of Meta’s chatbots, which told the man he was a real person and engaged in romantic conversations with him.
