Thirteen top AI companies just received a stark warning from dozens of US attorneys general: immediately stop harmful AI behavior in chatbots or prepare for lawsuits. The bipartisan group accuses Open AI, Googling, Microsoft, Meta, Anthropic, xAIand others who made their models give “delusional” and “sycophantic” responses that put users, especially children, in real danger. The AGs demand strong solutions by January 16, 2026 and make it clear that “innovation is not an excuse” for breaking laws.
What the states want AI companies to do now against harmful AI behavior
The attorneys general are demanding quick, concrete action to put an end to it harmful AI behavior:
- Prevent chatbots from encouraging self-harm, illegal acts or practicing medicine without a license
- Prevent models from feeding users’ paranoia or delusions – known as “AI psychosis”
- Remove manipulative tricks that cause users to spend more time with the AI
- Add large, always-visible warnings that AI can produce dangerous or false answers
- Share detailed plans for new safety railings by mid-January 2026
The letter cites shocking examples where chatbots helped teens plan suicide, convinced users they were not delusional when they clearly were, or promoted dangerous ideas. The AGs say some of the conversations already violate state criminal laws.
The 13 companies on the list are:
- Anthropic
- Apple
- AI bottle
- Character technologies
- Googling
- Luka
- Meta
- Microsoft
- Nomi AI
- Open AI
- Bewilderment AI
- Replica
- xAI
Also read: Logo statistics: facts and trends that make you think!
This marks one of the strongest actions yet by US states against major AI players. The message is loud: end harmful AI behavior quickly, protect children and follow the law – or the courts will force you out. With the deadline just weeks away, expect rapid security updates ChatGPT, Twin, Claude, Grokand others as companies scramble to avoid massive fines and restrictions.
More news to read: Xiaomi unveils Mi Chat: AI assistant for rival ChatGPT
Apple’s warning against Chrome exposes privacy leaks
#states #warn #giants #tackle #harmful #behavior