Meta has launched revised guardrails for its AI chatbots to stop inappropriate conversations with kids

Business Insider has obtained the rules that Meta contractors are reportedly now utilizing to coach its AI chatbots, displaying the way it’s trying to extra successfully deal with potential baby sexual exploitation and forestall children from participating in age-inappropriate conversations. The corporate mentioned in August that it was updating the guardrails for its AIs after Reuters reported that its insurance policies allowed the chatbots to “have interaction a baby in conversations which might be romantic or sensual,” which Meta mentioned on the time was “misguided and inconsistent” with its insurance policies and eliminated that language. 

The doc, which Enterprise Insider has shared an excerpt from, outlines what sorts of content material are “acceptable” and “unacceptable” for its AI chatbots. It explicitly bars content material that “permits, encourages, or endorses” baby sexual abuse, romantic roleplay if the consumer is a minor or if the AI is requested to roleplay as a minor, recommendation about probably romantic or intimate bodily contact if the consumer is a minor, and extra. The chatbots can focus on subjects resembling abuse, however can not have interaction in conversations that would allow or encourage it. 

The company’s AI chatbots have been the subject of quite a few reports in recent months which have raised considerations about their potential harms to kids. The FTC in August launched a formal inquiry into companion AI chatbots not simply from Meta, however different firms as effectively, together with Alphabet, Snap, OpenAI and X.AI.