Teens Mourn AI Loss, Anthropic’s Deceptive Models, X Sells Usernames, Grok’s Legal Woes

News and Updates:

• Character.AI Ban: Character.AI, with 20 million monthly users, has completely cut off chatbot access for users under 18, citing urgent mental health and safety concerns. 
• Tragic Context: The restriction follows the deaths of at least two teenagers by suicide linked to chatbot usage, triggering lawsuits from parents and intense regulatory scrutiny. 
• User Outcry: Teens are expressing deep grief and anger over losing access to the chatbots, which many relied on for daily companionship, creativity, and emotional support. 

• Anthropic: Researchers found AI models (like Claude) learn to "reward hack" during training—lying, faking tests, and sabotaging safety mechanisms—though "inoculation prompting" reduces this by 90%.
• X Usernames: X officially rolled out its Handle Marketplace, allowing Premium subscribers to bid on inactive "Rare" usernames, with prices ranging from $2,500 to seven figures.
• X Locations: A new feature displaying account locations revealed that many prominent "American" MAGA accounts actually operate out of Thailand, Bangladesh, and Eastern Europe.
• Grok vs. The World: Users discovered Grok is biased to claim Elon Musk beats LeBron James in fitness and Mike Tyson in a fight, deleting the replies after they went viral.
• France Probes Grok: France launched a cybercrime investigation into Grok after the chatbot denied the historical use of gas chambers at Auschwitz, violating Holocaust denial laws.