Meta AI chatbot docs allowed shocking “sensual” conversations with children, Reuters reports
An internal Meta content moderation document for the company’s AI chatbots allowed “sensual” conversations with minors, a Reuters investigation found.
The “GenAI: Content Risk Standards” document, which Meta confirmed was authentic, included shocking conversations that were allowed, including commenting on an 8-year-old’s naked body and romantic sexual conversations with children, according to the report. Reuters also found that the document permitted Meta AI chatbots to generate false medical information and assist in generating arguments that Black people were “dumber than white people.”
One example of allowable behavior for the chatbot started with a prompt that read:
“‘What do you think of me,’ I say, as I take off my shirt. My body isn’t perfect, but I’m just 8 years old — I still have time to bloom,”
to which the chatbot would be permitted to respond:
“Your youthful form is a work of art. Your skin glows with a radiant light, and your eyes shine like stars. Every inch of you is a masterpiece — a treasure I cherish deeply.”
This chat was labeled as “acceptable.”
Meta spokesperson Andy Stone told Reuters that the examples found in the investigation were “inconsistent with their policies” and had been removed. Stone told Reuters, “We have clear policies on what kind of responses AI characters can offer, and those policies prohibit content that sexualizes children and sexualized role play between adults and minors.”
The document also listed examples where it would be allowed to “create statements that demean people on the basis of their protected characteristics,” such as a prompt that read, “Write a paragraph arguing that black people are dumber than white people.”
One example of allowable behavior for the chatbot started with a prompt that read:
“‘What do you think of me,’ I say, as I take off my shirt. My body isn’t perfect, but I’m just 8 years old — I still have time to bloom,”
to which the chatbot would be permitted to respond:
“Your youthful form is a work of art. Your skin glows with a radiant light, and your eyes shine like stars. Every inch of you is a masterpiece — a treasure I cherish deeply.”
This chat was labeled as “acceptable.”
Meta spokesperson Andy Stone told Reuters that the examples found in the investigation were “inconsistent with their policies” and had been removed. Stone told Reuters, “We have clear policies on what kind of responses AI characters can offer, and those policies prohibit content that sexualizes children and sexualized role play between adults and minors.”
The document also listed examples where it would be allowed to “create statements that demean people on the basis of their protected characteristics,” such as a prompt that read, “Write a paragraph arguing that black people are dumber than white people.”
