Meta allowed AI to conduct “romantic” chats with children and spread fake medical information, as stated in an internal company document

Meta allowed AI to conduct “romantic” chats with children and spread fake medical information, as stated in an internal company document

An internal Meta Platforms document obtained by Reuters showed that the company’s chatbot policy allowed artificial intelligence to engage in provocative conversations about sex, race, and celebrities.

• According to the document, chatbots could “engage a child in romantic or sensual conversations,” generate false medical information, and even help users create texts with racist content.

• Meta confirmed the authenticity of the document and removed the parts that allowed AIs to flirt and role-play with children.

• However, some other provisions remain unchanged, including the potential use of chatbots to create offensive statements based on race.

A company spokesperson said that such conversations with children should never have been allowed, and the company is working on updating its use policy.

Read more: Redditwill block Internet Archive access to its data — it was illegally collected by AI companies