Bankruptcy law was not built for chat logs, experts say of Character.AI case

Character.AI's Chapter 11 case is testing whether bankruptcy procedure can cope with intimate conversational data, according to a Bloomberg Law analysis published on 12 November. The outlet describes it as the first bankruptcy of a consumer-facing chatbot company.

What we know

Bloomberg Law reported that the company's chat logs span more than 20 million users, among them over 2 million accounts held by minors. Disney's $20 million stalking-horse offer covers a licence to certain intellectual property and technology and does not include those logs.

Lev Breydo, a law professor at William & Mary, told the outlet that three kinds of data are in play: the conversations themselves, transaction records, and model weights trained on the conversations. Existing bankruptcy tools, he said, were designed with customer lists in mind. Nizan Packin of Baruch College made a similar point about the consumer privacy ombudsman, a role conceived for names, emails and birthdays and not for sensitive exchanges involving children. The judge has indicated that Character.AI should appoint such an ombudsman for the Disney deal, Bloomberg Law said.

The data cannot simply be discarded either. With state attorneys general investigating, the logs may be evidence, the academics noted. Ben Winters of the Consumer Federation of America called the situation "one of the, if not the biggest nightmare scenario," while Calli Schroeder of the Electronic Privacy Information Center said users' protections remain unclear. The article points to last year's $305 million sale of 23andMe's genetic data as the nearest precedent.

Why it matters

The model-weights point deserves attention. Even if raw logs are never transferred, a system trained on them carries their imprint, and no established rule says whether that counts as selling personal data. A buyer of technology alone may still inherit a privacy question.

For operators, the case shows that insolvency is now a data-governance scenario to plan for. Retention limits, separation of training data and a clear position on what happens to conversations in a sale would all be easier to settle before a court has to. For investors, chat archives look less like an asset and more like a contested liability.

What to watch

The ombudsman's appointment and findings, and whether any bid before the 24 November deadline reaches for the consumer platform or its data.