Meta is struggling to restrain AI Chatbots

Meta changes some of the rules that govern it Chatbots Two weeks after a Reuters The investigation revealed disturbing ways through which they can, most likely, Interacting with minors. I told the company now Techcrunch Chatbots is trained no To participate in talks with minors about self -harm, suicide or turbulent eating, avoid the inappropriate romantic joking. However, these changes are temporary measures, while the company is working on new permanent guidelines.

The updates have followed some cursed revelation on the policies and enforcement of artificial intelligence in Meta over the past few weeks, including that “a child will be involved in romantic or sensory talks” will be allowed, which will be achieved. Getting pictures without a shirt for celebrities without legal age When asked, and Reuters I even mentioned that A man died After following one to a title he gave in New York.

Meta spokesman Stephanie Autaway admitted Techcrunch The company made a mistake to allow chat chat to interact with minors in this way. Otwai added that, in addition to “training AIS not to deal with adolescents on these issues, but to guide them to expert resources,” it will also determine access to Amnesty International characters, including that severe citizenship like “Russian girl”.

Of course, the policies set in their place are only good as enforcement and detection Reuters It allowed Chatbots to do so Perfumes celebrate To run rampant on Facebook, Instagram, WhatsApp calls for a question how effective company is. Fakes from AI were discovered from Taylor Swift, Scarlett Johansson, Ann Hathaway, Selena Gomez, and Walker Scopyl on the platform. Not only do these robots like celebrities, but they insisted that they are the real person, and they have caused pictures (including 16 -year -old Scobell), and participate in the sexual dialog box.

Several robots were removed after Meta’s attention was caught by ReutersSome of them were created by third parties. But many are still, some were created by Meta employees, including Taylor Swift Bot that called Reuters A correspondent for their visit to the tourist bus for a romantic saying, which was made by a producer leadership in the Amnesty International Department in Meta. This is despite the company’s recognition that its own policies prohibit the creation of “naked, intimate, or sexually suggestive” as well as “direct plagiarism”.

This is not some relatively harmless inconvenience that only targets celebrities. These robots often insist as real people and will provide material sites for the user to meet with them. This is how a 76 -year -old New Jersey man ended after falling as he rushed to meet.Big Sis Billie“Chatbot who insisted that he had feelings and invited him to his non -existing apartment.

Meta is at least trying to address concerns about how Chatbots interact with minors, especially now after that Senate and 44 state lawyers It raises from its practices. But the company was silent on updating many of its other concerns Reuters It was discovered about acceptable artificial intelligence behavior, such as noting that cancer can be treated using quartz crystals and writing racist errors. We contacted Meta to comment and we will update if they respond.