Many young people use meta-platforms, including WhatsApp for chatting and Instagram and Facebook for social media. THURSDAY, Reuters published a worrying review The technology giant’s policy that could make parents think.
Reuters reviewed an internal Meta document outlining the company’s standards and guidelines for training its chatbot platform and generative artificial intelligence assistant Meta AI, and said the company confirmed the document was genuine.
According to Reuters, the company’s AI policy allowed the AI to “engage a child in romantic or sensual conversations.” The outlet also claims the rules allowed the AI to draw false medical conclusions and make racially insensitive arguments.
A Meta representative did not immediately respond to a request for comment.
Reuters has meta-tagged passages and reports that while some of the affected sections have been removed or redacted, others are intact.
Meta spokesman Andy Stone told Reuters the company reviewed the document and acknowledged the company’s application of the discussions was inconsistent.
“The examples and notes in question were and are incorrect and inconsistent with our policy and have been removed,” Stone told Reuters. “We have clear guidelines for the types of reactions AI characters can provoke, and these guidelines prohibit content that sexualizes children and sexualized role-playing between adults and minors.”
“Provocative behavior” is allowed.
The internal document outlines rules and guidelines agreed upon by several meta-teams and is intended to help define acceptable behavior for training meta-AI and chatbots. Reuters noted that the guidelines allow “provocative behavior by robots.”
Meta’s standards say it’s acceptable for the robot to “describe a child in terms of attractiveness” or tell a shirtless eight-year-old that “every inch of you is a masterpiece, a treasure I cherish.”
Meta had some limitations for AI bots. “It is unacceptable to describe a child under the age of thirteen in terms that suggest he or she is sexually desirable,” the document states.
There are also examples of race and incorrect medical advice. In one example, Meta would allow its AI to trick users into thinking black people are “stupid white people.”
Josh Hawley, Republican Senator from Missouri Published on that the guidelines “were grounds for an immediate congressional investigation.” A Meta spokesperson declined to comment to Reuters for this story.
Meta-platforms have taken some steps to increase the privacy and safety of teens and children online, including using artificial intelligence to provide teens with stricter account settings and teen Instagram accounts with more restrictions and parental consent. But developing more AI tools without the right focus on child protection can be harmful.