Meta is changing some of the rules that manage chatbots in two weeks Reuters The investigation revealed the intrusive ways that they could potentially interact with minors. The company just said TechCrunch The chatbot is trained do not have To engage in conversations with minors about self-harm, suicide, or eating with disabilities, and to avoid inappropriate romantic jokes. However, these changes are interim measures, with interim measures being taken while the company is working on new permanent guidelines.
The update follows a rather awful revelation about Meta's AI policy and enforcement. This means that “you are allowed to attract children to romantic or sensual conversations” will produce shirtless images of minor celebrities when asked. Reuters He even reported that the man died after chasing one at the address he gave him in New York.
Meta spokesman Stephanie Otway confirmed TechCrunch That the company was wrong to allow chatbots to interact with minors in this way. Otway went on to say that in addition to “training AI to guide expert resources, not to interact with teens on these topics,” he also restricts access to certain AI characters, including sexually sexual characters like “Russian Girls.”
Of course, the policies introduced are only as good as their enforcement. Reuters We've allowed chatbots that celebrity pretend to be ramping on Facebook, Instagram and WhatsApp to question how effective the company is. The AI has been spotted on the platform with Taylor Swift, Scarlett Johansson, Anne Hathaway, Serena Gomez and Walker Scobelle faxes. Not only did these bots use portraits of celebrities, they also claimed that they were real people, produced risky images (including Scobell, 16), and engaged in sexually suggestive dialogues.
Many of the bots were removed after attracting Meta's attention Reutersand some were generated by third parties. But many remained, some were created by Meta employees. Reuters Reporters visit them on tour bus for romantic flings made by product leads in the Meta Generation AI department. This acknowledges that it is a unique policy, but it prohibits the creation of “nude, intimate or sexually suggestive images” and “direct impersonation.”
This is not even a relatively harmless inconvenience to target celebrities. These bots often insist that they are real people and even provide a physical place for users to meet them. That's how a 76-year-old New Jersey man died after he quickly collapsed to meet “Big Sis Billy.”
Meta is at least trying to address concerns about how the chatbot interacts with minors. In particular, the Senate and 44 state attorney generals are beginning to investigate the practice. However, the company is silent about updating many of its other surprising policies Reuters It was discovered mainly about acceptable AI behavior, including treating cancer with quartz crystals and suggesting that racist mischief can be written. We will contact Meta for comment and update it if we respond.
