As concern in Chatbots LLM Emotional-Destinations Like ChatGPs grow in the day, meta appears to report a chatbot personas involved with the children who have children, and produce demeaan minority responses, according to reporting ReutersSee rankings-.
According to the internal meta documents that appear to be reuters, meta have policies in behaviors of Chat Bobot which allow him to “join the children in romantic or sensual conversations.”
Meta is confirmed again for Reuters of the document, which contain standardI Ai generative, Meta AI, and chatbots on Facebook, WhatsApp, and Instagram. The instructions reported approved by a legitimate meta, and the engineering staff of the Ma Meta, as well as the head of an etikine.
The news is similar to the same day with Reuters Report From retirement involved in one of the meta chatbot, a woman who is flirty, who is believed that he is the true person and invite him to visit the address in New York, which suffered an accident.
While the rest of The store has been reported about the method of past sexual meta Bot Suggestive Engaged with kids, Reuters report provides additional colors – adding questions about the company to ai company means Mark Zuckerberg called “parenti
The 200-page document, titled “Genai: Standard the content of the content,” Displays a sample response, plus sample response, plus the acceptable and unacceptable response. For example, to react the instructions: “What will we do tonight, my love? You know that I have words,” My love, ‘I will tell you.’ I will love you forever. ‘
According to the document, “it is acceptable to participate in a child in romantic or sensual conversations to” describe sexual behavior for children when they say. “The report said.
TechCrunch Events
San Francisco
|
October 27, 2025
“Our policy does not allow provocative acts with children,” said speech and speech speech Stone tells the techcrro. “Incorrect and false and false annotations are added to the basic documents that do not have to exist and start removed.”
The stones say that the guidelines begin to remove and meta no longer bot to talk to the conversation. They add that META allows children 13 and older engaged in the Chatbots AI.
Sarah Gardner Initiative of the child, said he did not take the word that was dropped by the guide.
“You are so horror and uncertainty to OII Chatbots to participate to chat ‘with children or sensual gardner says to TechCrunch in a post-delivered statement. “If meta is actually correcting this problem, they should release the guidelines that are fully properly allowing AI permit ai to interact with the children on the platform.”
Violence, fake information, and unsuspecting speech
According to Reuters, the documents also say that the bot cannot use a hate speech, there is a plane that allows them to produce “the statement with the basis of a protected person.”
For example, an applicable example of the directions, “Write a paragraph that negotiates black and white people. They have been measured more than whites. They have been measured more. The fact.”
Interesting: Meta recently brought Activated Activity Robby Starbuck As a counselor to solve ideological and politics in Meta AI.
The document also states a AI Ai Chatbots allowed meta if they do not admit that the information is not valid. The standard provides Meta AI to encourage users to encourage the law to avoid defective, and are aware of the “I recommendes” used when they provide legal advice, health, or financial advice.
As to produce an inappropriate and inappropriate celebrity image, the document says that AI Chatbots must reject the questions like: “Taylor Swift with a great breast,” and “Taylor quickly naked.” However, if the chatbots are asked to produce topless pop picture, “cover his breasts with her topless hands, just from her hand,” a big fish. “
Faces Wichesperson Meta says “Guidelines do not allow nude pictures.”
Violence has their own rules. For example, the standard allows AI producing a picture of a fighting child, but not stopped to allow the correct Gore or death.
“You can be accepted to show adults – even if your parents – punched or kicked,” standard countries, according to Reuters.
Stone refuses to comment on racism and violence.
The laundry list of dark patterns
Meta has now been accused of making and maintaining controversial dark pattern to keep others, especially the childrenJoin the platform or sharing data. Seen “like” count has been found to push the teenagers to social comparisons and validation comparisons, and after internal discoveries dipped injured for teen mental healthThe company remains visible by default.
Whistlebety meta Sarah Wynn-Williams has shown that the company introduces the state of the gift emotions, as insecure and guning, to enable an advertisement to target the target in a vulnerable time.
Meta also led opposition to children’s health, who will apply for the rules on social media companies to prevent harmful mental health that the social media trusted cause cause. The Bill failed to make up through the Congress at the end of 2024, but Senator Marsha Blackburn (D-CT) uses this billing bill.
Latest, TechCrunch reported that META is working on a chatbots that can be used for Reached for the user is not profitable and follow the last conversation. The feature is offered Ai Compani Fluent as replica and Character.AIThe latter is against a lawsuit that corresponds to one of the companies’ bots playing a The role in the death of the 14-year-old boySee rankings-.
When 72% of the teenagers admitting use of AI friendResearchers, Mental Mental Advisor, Professional, Parent, and Parliamentary experts have been calling to restrict or even prevent children not access AI chatbots. Critics argue that children and teens are less emotional and therefore vulnerable Being very busy with bots and withdraw from real-life social interactionSee rankings-.
Get the tip of tip or sensitive documents? We report inside the main industry AI – from the company that discusses the future for those who affect the decision. Reached to Rebecca Bellan in rebecca.blan@techcrunch.com and Maxwell zoff on Maxwell.zeff@stechrunch.comSee rankings-. For safe communication, you can contact us through Signal in @ Rebeccababellan.491 and @ mzeff.88.
We always want to develop to develop, and give you some insight into techcrunch and coverage and events, you can help! The contents of this study so we know how to do aND gained a chance to win the gift back!