Character.AI Bankruptcy Puts Millions of User Chats on the Auction Block

Character.AI filed for bankruptcy this month and opened a court-supervised sale of its assets, and nobody knows how much data on its millions of account holders will be sold with the company. The most valuable asset on offer may be the company's LLM (large language model, the core AI system that generates responses), which users have substantially trained themselves while building and interacting with characters on the platform.

Legally, a company can sell personal data about its users if its privacy policy permits the transfer. Character.AI's privacy policy is vague. It says the company may disclose "certain information" in a sale. Nobody knows what that means.

The US Trustee Office, an arm of the Justice Department, wants the presiding judge to appoint a consumer privacy ombudsman to advise the court on user privacy. The judge could restrict the sale of company "records," including user data.

Why does bankruptcy law not protect the model?

The Bankruptcy Code has no provision specifically addressing trained models. The legal status of LLMs and their training data has yet to be settled in law. That means the data users fed into Character.AI could be sold to whoever buys it.

Character.AI is one example. On that service, users form "relationships" with customizable characters, including ones they create and share with the community. Some users chat about thoughts of suicide and self-harm, and they surrender mental health disclosures, ages, private chats, voice calls, deep fears and relationship struggles.

User awareness is low. A survey commissioned by DuckDuckGo and published in August asked 1,944 American adults about chatbot privacy. More than half, 53 percent, did not know whether their conversations were being used to train the bots. They are. Only a quarter knew chatbot conversations can be subpoenaed. They can. Among self-described AI enthusiasts, 56 percent said they had confided to a chatbot something they had never told a close friend, parent, colleague, doctor or therapist.

Can shared chat data end up inside an LLM?

The answer is "maybe." Major chatbot companies vary widely in how they handle chat data, and the differences boil down to four questions: whether training on chats is on by default, whether human reviewers read them, how long they are kept, and whether they feed advertising.

The companies publish privacy controls, but their own policies admit consumer chats may be used for training by default, read by human reviewers, or both. Some offer enterprise or business accounts that handle chat sessions privately and exclude them from training data — an implicit admission that individual accounts are not private. OpenAI lets users opt out of having their chats used as training data, but its help documentation states that turning off training does not delete the chats themselves.

There are three main exposure vectors:

• Bugs. In March 2023, a bug in the open-source library redis-py let some ChatGPT users see other users' chat titles and possibly the first message of new conversations. More than 1 percent of ChatGPT Plus subscribers active in a nine-hour window may have had payment details exposed: names, email addresses, payment addresses, card types, the last four digits of card numbers and expiration dates.
• Sharing features. In 2025, Forbes reported that Google results showed more than 370,000 Grok conversations after users clicked "share." Shared ChatGPT chats turned up in search results until OpenAI killed the option.
• Absorption into training data. In a 2022 study, GPT-Neo models given only a person's name recovered the right email address no more than 1.24 percent of the time. With the domain supplied, the rate climbed to 16.55 percent, and to 37.06 percent after five examples.

Does deleting chats fix anything?

No. Several state attorneys general said deleting chat records may not undo their influence on trained models. Robbie Torney of Common Sense Media claimed that exporting or deleting data "may not mean that that data is gone from model training."

Users also enter information about co-workers, friends and relatives, not only about themselves. The main thing standing between those secrets and the world is company practice.

What happens when AI companies fail?

Talk of an "AI bubble" bursting is constant. In China, it has already burst, according to Reuters. Shanghai's STAR50 index of tech stocks fell 40 percent from its July high in under four months, erasing about $480 billion in market value. The broader CSI300 is down about 20 percent from its June highs. That is not a total collapse, but it pressures US stock prices and creates incentives to transfer ownership through sales or mergers, which has not yet happened in China.

Which AI companies will go bust as Character.AI did, get sold off, or get hacked is unknowable. The chatbots' inner workings are equally unknowable. The conclusion for enterprises and individuals is the same: never put sensitive information into chatbots.