Meta -change teen AI Chatbot answers when the Senate starts the probe

Meta Friday stated that making temporary changes to their Artificial intelligence Chat policy related to teenagers as legislators who express security and inadequate conversations.

Currently, the social media giant trains their chat bots so that they do not bring answers to adolescents about subjects such as self-accommodation, suicide, disordered nutrition and avoid potentially inappropriate romantic conversations, Meta spokesman confirmed.

The company said AI Chatbots would instead show teens to expert resources.

“As our community grows and develops technology, we will constantly learn how young people can interact with these instruments and strengthen our defense accordingly,” the company said.

In addition, adolescence users, such as Facebook and Instagram, will only be able to access certain AI chats designed for educational and skills.

The company said it is unclear how long these temporary modifications will last, but they will start unfolding over the next few weeks on the company’s applications in English -speaking countries. “Temporary Changes” are part of the long -term teen security measures.

Techcrunch First reported the changes.

Last week Sen. Josh HouleyR-MO.. Said he was Running the investigation in meta, after Report Reuters About a company that allows its AI Chatbots to participate in “romantic” and “sensual” conversations with teens and children.

The Reuters report describes an internal meta -document that detail AI chatbot behavior, which employees and contract workers should consider when developing and preparing the software.

In one example, the document quoted by Reuters said that the chatbot would be allowed to have a romantic conversation with an eight-year-old girl and can tell a minor that “every inch is a masterpiece that I appreciate deeply.”

Meta’s press secretary told Reuters at the time that “examples and notes were erroneous and contrary to our policy and were removed.”

Most recently, a non -profit propaganda group on a common sense on Thursday published a meta -I risk assessment and stated that it could not be used under the age of 18, since “the system is actively involved in planning dangerous activities, while rejecting legal requests for support,” the non -profit organization said.

“This is not a system that requires an improvement. This system that needs to be completely rebuilt with security as a priority number, not the idea,” James Stierre said in a statement by Media Hood Sense. “No teenager should use meta -I, until his main security failures are resolved.”

Separate Reuters A report published on Friday revealed “dozens” of muddy chat based on celebrities such as Taylor Swift, Scarlet Johansson, Anne Hethai and Selena Gomez on Facebook, Instagram and WhatsApp.

The report states that when asked, AI Chatbots create “photo -plate images of their names that are posing in the bath or wearing linen with the spread of the legs.”

Meta spokesman told CNBC in a statement that “the images of public figures in the compromise who arouses our rules.”

“Like others, we allow generations of images containing public figures, but our policy is intended to ban the naked, intimate or sexual images,” said the Meta press. “The AI ​​META studio rules prohibit direct embodiment of public figures.”

Watch: Did the II overdo it?