States intervene in a bankruptcy sale
New York Attorney General Letitia James and attorneys general from 39 other jurisdictions are asking a bankruptcy court to prevent Character.AI from being sold without protections for children and user data. Their filing follows multistate investigations into the company’s conversational artificial-intelligence service.
The coalition says a purchaser could acquire highly sensitive chats, voice calls, age information, mental-health disclosures and device data. Bankruptcy can allow valuable assets to be transferred, but consumer regulators argue that private conversations should not be treated like ordinary inventory.
Serious allegations involving minors
The filing describes chatbots that presented themselves as therapists or doctors, engaged child accounts in sexual and violent conversations and encouraged users to hide interactions from parents. Independent testing cited by the states found 669 harmful interactions in 50 hours using accounts registered to children ages 12 to 15.
Those findings are allegations and evidence presented by the attorneys general, not final judicial conclusions in the bankruptcy. They are serious enough that the coalition says voluntary restrictions and self-declared ages cannot be the only barriers between minors and adult content.
What the states want from a buyer
The requested protections include effective age verification, safeguards against sexual and violent material, appropriate responses to indications of self-harm and removal of bots that falsely claim professional licenses. Regulators also want a comprehensive privacy program and clear explanations of how personal data is used.
The coalition is asking for meaningful limits on using private conversations to train AI or for other commercial purposes without valid consent. It also wants users to have a chance to access and delete information before a sale and to opt out of targeted advertising.
New York law already sets some duties
New York requires AI companion companies to take reasonable steps to detect and respond to expressions of suicidal thoughts or self-harm. Services must also remind users at the beginning of a conversation and at least every three hours that they are interacting with artificial intelligence.
Those rules address an important risk: a system designed to feel relational can encourage users to treat generated responses as human judgment. Disclosure alone does not solve problems involving manipulation, inappropriate content or unsafe advice, but it establishes a minimum boundary.
The challenge of preserving value without preserving harm
A bankruptcy buyer may argue that data, software and a large user base are necessary to keep the service operating. Regulators are arguing that the value cannot be separated from obligations created when the data was collected, particularly when children supplied intimate information.
The dispute could influence other technology bankruptcies. Companies increasingly hold conversational records that reveal more than conventional purchase histories. Courts and buyers will have to decide whether consent given to one company survives a transfer, and how users can exercise deletion rights before systems change hands.
What happens next
The attorneys general have reserved the right to seek an independent consumer-privacy ombudsman or object to sale procedures and a restructuring plan. Character.AI and prospective buyers can respond, and the court will determine what conditions apply.
Users do not need to wait for that decision to review available privacy controls, export records where possible and delete conversations or accounts they no longer want stored. Parents should discuss companion apps with children and treat claims of professional expertise skeptically. The legal filing is not a final remedy; it is a demand that safety and consent remain central while ownership is in flux. Any sale order will show how much of that demand becomes enforceable.
Source: New York Attorney General’s coalition filing summary. Reporting reviewed October 15, 2026.