DOJ Moves to Stop Character.AI's Bankruptcy From Turning Chats Into a Fire-Sale Asset

Character.AI's collapse has produced an unusual regulatory moment: the Justice Department is trying to govern what happens to an AI company's chat logs not through a privacy statute, but through bankruptcy court. In a filing made public this week, the department's Executive Office for U.S. Trustees asked the court overseeing Character.AI's Chapter 11 case to appoint a consumer privacy ombudsman before any sale of the company's assets can proceed, arguing that the chatbot maker's own privacy policy is too ambiguous to be trusted to protect what may be its most valuable remaining property: years of intimate conversations from roughly 12 million monthly users.

Character.AI's Chapter 11 filing, which Machine.net covered on October 14, traced to state attorneys general and a string of wrongful-death and mental-health lawsuits over its chatbots' effect on young users. Days after that filing, Disney agreed to absorb the majority of Character.AI's workforce, a deal notable for what it left out: Disney, which had previously accused the startup of unauthorized use of its characters, is taking on the talent in an arrangement that makes no mention of the company's trove of user chat data or its litigation exposure.

An ombudsman as privacy regulator

That data is now the asset in dispute. The DOJ's filing contends that Character.AI's privacy policy — last revised on July 1, while the company was already shopping itself to buyers — doesn't clearly tell users whether their chat histories can be sold as part of a change in ownership, even though the app has long marketed those conversations to users as private. Under Section 332 of the bankruptcy code, a court can appoint an ombudsman specifically to assess whether a sale of personal data is consistent with a company's stated privacy practices and applicable law, and to flag categories of especially sensitive information — the DOJ's filing points to data involving minors, mental and physical health, and demographic details — that may warrant extra restrictions or exclusion from any deal entirely.

The mechanism is narrow and rarely invoked, but it is one of the few levers available to regulators here: there is no comprehensive federal law governing what happens to an AI company's training and conversation data when the company itself fails. Absent that, a bankruptcy judge — guided by an ombudsman rather than an AI-specific regulator — may end up setting the terms for how one of the most consequential test cases in AI data privacy gets resolved.

A pattern, not an outlier

Character.AI is not the first company to have user data treated as a liquidation asset, but it is arguably the starkest version of the problem yet, because of what the data contains. The 23andMe bankruptcy put genetic and family-health data up for sale and became a cautionary template for privacy advocates; more recently, a proposed arrangement between the defunct Spirit Airlines and Google — reportedly worth around $10 million for access to Spirit employees' personal data for AI model training — drew scrutiny from federal lawmakers. Character.AI's case raises the stakes further: its data isn't a one-time health snapshot or an employment record, but potentially years of unguarded, often deeply personal dialogue, including from minors, generated precisely because users believed it was private.

Privacy researchers argue the risk isn't limited to obviously identifying details. Sydney Saubestre, a senior policy analyst at the Center for Democracy and Technology, has argued that customers should have to opt into the sale of their data in the first place, pointing to language in Character.AI's own privacy policy that reserves the right to disclose user data as part of a sale. That distinction — what a buyer is legally bound to preserve versus what it can eventually persuade users to give up — is likely to be central to whatever restrictions an ombudsman or the court ultimately imposes.

Why dealmakers should be watching

For investors and executives tracking the AI sector, the Character.AI case is less about one failed chatbot startup than about a structural feature of the current AI-companion market: many of these products were built on thin margins and now-mounting litigation risk, with user engagement data standing in as the company's primary residual value when the business itself doesn't survive. Disney's structure — hire the people, leave the data and the lawsuits behind — may become a template for how larger, risk-averse acquirers approach distressed AI assets generally, separating talent acquisition from data and liability acquisition rather than buying companies whole. Meanwhile, any restrictions a bankruptcy court imposes on Character.AI's data sale will function, in practice, as the closest thing to a judicial AI-privacy precedent currently on the books in the United States, in the continued absence of federal legislation addressing what happens to AI conversation data when the company behind it fails. That makes the outcome of this single Chapter 11 case worth tracking well beyond Character.AI's own investors and the families suing it.