The rapid growth of the artificial intelligence industry carries a risk that is rarely considered: if an AI company goes bankrupt, the personal data databases it has accumulated become an asset that can be sold to third parties. This is the conclusion reached by journalist Thomas Germain in a BBC Future article that illustrates the threat mechanism using the Character.ai service as an example.

What Happened

On October 29, 2026, BBC Future published an article by Thomas Germain titled "The AI bubble could expose your deepest secrets," focusing on the fate of user data in the event of AI company bankruptcies. The central case is Character.ai: the company sold part of its assets to Disney for 20 million dollars, but user data was not included in the deal. Courts have indicated that Character.ai has the right to sell the accumulated correspondence to other buyers, meaning the conversational database is legally separable from the service itself. According to Character.ai's privacy policy, this data can be used for model training, profiling, and targeted advertising, so the value of the asset is not limited to the text of the conversations themselves.

Context

The specific nature of the situation lies in the structure of US legislation: in bankruptcy, the transfer of data to buyers is often explicitly written as an exception to data protection promises, so the privacy policy itself does not cancel such a scenario. Experts interviewed by BBC Future—Epic lawyer Calli Schroeder and UNC Chapel Hill professor Melissa Jacoby—draw attention to the nature of conversational data: with "personified" bots, users share more sensitive information, including details about friends and relatives who never used the service and did not consent to data processing. As a result, a bankrupt companion bot service leaves behind a potentially sellable array where a significant portion of the mentioned individuals were not even parties to the agreement.

Why This Matters for the Industry

For the industry, this is a story about data management as much as it is about privacy: if you fine-tune models on raw user dialogs, you are creating an asset that will outlive the product itself. The Character.ai precedent changes the rules of deals: buyers of AI company assets must assess not only the technology and team but also the fate of user databases, while working consumer services face retention pressure—learning about the right to resell data, people reduce self-disclosure, even though openness with a bot is a key value of such products. Practical steps for today include auditing derivative artifacts of user data (logs, caches, eval sets, training corpora), removing personal information (including mentions of third parties) before any fine-tuning, and aligning product promises with the actual deletion pipeline. If the wave of AI company consolidation continues, the fate of conversational datasets may become a standard due diligence item in deals.

Why This Matters for Users

The risk is personal: if you have communicated with AI bots, especially "personified" ones, your messages and even mentions of your loved ones could end up with a new data owner after the developer's bankruptcy. A concrete action is to review the services where you shared sensitive information and delete your account along with the data, after first studying the deletion instructions in the specific product's privacy policy, as there are no unified standards here.

What Is Still Unknown / Limitations

The BBC Future material is a journalistic analysis featuring two experts and one case: it contains no quantitative assessment of the scale or statistics on AI company bankruptcies, so speculations about a wave of bankruptcies and the formation of a resale market for conversational databases remain hypotheses requiring verification. Product promises of "privacy by architecture"—local user-side storage, ephemeral chats, one-click deletion—are also debatable; the source provides no methodology for verifying such claims, and "verifiable deletion" without an audit protocol and metrics remains a declaration.

Sources

Author

Look at AI, editorial team