ings
On October 12, 2026, Character Technologies, the company behind Character.AI, filed for Chapter 11 protection in Delaware. It plans to sell all or substantially all of its assets under court supervision while keeping the platform running during the process. The opening offer came from Disney Worldwide Services: $20 million for a package involving a non-exclusive, perpetual license to certain intellectual property and certain company legal claims, with transition services provided by former Character.AI employees now working at Disney.
Just over two years earlier, Google reportedly struck a deal worth roughly $2.7 billion with the company, obtaining a technology license as its two founders returned to Google. Comparing $2.7 billion with $20 million can suggest a dramatic collapse in value. Yet the transactions differ in scope, terms, and circumstances. They should not be treated as two valuations of the entire company.
The more revealing question is why Character.AI’s extensive usage did not generate sufficient revenue to sustain the business. Public accounts have frequently put the platform at around 20 million monthly active users and roughly 75 minutes of daily use. If its technology can attract a bid and its people can find new employers, why is the platform—with its characters and accumulated conversations—still waiting for someone to take it forward?
01 Users Spend Time. The Company Pays to Serve It.
Character.AI’s appeal is easy to understand. Users can choose an existing community character or create one by defining its identity, personality, and speaking style, then shaping it through sample dialogue. Fictional figures, public personalities, and original creations can all become conversation partners. As those conversations continue, users accumulate more than messages: they develop an ongoing experience with a particular character.
The product later expanded beyond text into voice calls, group chats, character animations, interactive books, short series, and comics. Its early growth was rapid. The web beta launched in September 2022; around six months later, the company raised $150 million in a Series A round at a $1 billion valuation. Users had already created roughly 2.7 million characters. By August 2023, the company reportedly still employed only about 35 people.
Usage and revenue, however, did not grow in step. Public materials from different periods put monthly active users at around 20 million and average daily use at roughly 75 minutes. Annualized revenue around mid-2025 was approximately $30 million. A separate media estimate for late 2024 put paying subscribers at about 139,000—less than 1% of a roughly 20 million monthly active user base. The subscription price at that time was $9.99 per month.
That same estimate placed monthly operating costs at approximately $30 million. These figures come from different periods and measurement methods, so they cannot establish an exact loss. They nevertheless illustrate the pressure: spending could reach in a month what revenue generated over a year.
This is a central difficulty for AI companion products. Long sessions show that a product holds people’s attention, but serving those sessions requires ongoing inference capacity and operational resources. If the revenue generated by another hour of use does not cover the cost of providing it, stronger engagement can deepen financial pressure. Seventy-five minutes demonstrates appeal; it does not, by itself, demonstrate profitability.
02 August 2024 Changed the Company’s Center of Gravity
Much of Character.AI’s original promise came from its founders. Noam Shazeer co-authored the Transformer paper, while Daniel De Freitas started the conversational AI project at Google that later became LaMDA. They left Google in late 2021 to build a company that would develop both the underlying models and the consumer experience.
That structure changed in August 2024. Google obtained a non-exclusive license to Character.AI’s technology, and both founders joined Google along with a group of researchers. Character.AI continued operating independently, with general counsel Dominic Perella becoming interim CEO and the business placing greater emphasis on product development and user growth.
There is an important distinction between reducing investment in foundation models and losing all model-development capability. The company’s stated direction at the time was to make greater use of third-party models alongside its own, freeing resources for post-training and product experiences. That does not establish that every subsequent conversation used a purchased model, or that every message incurred a payment to an outside model provider. A contemporaneous quotation of the company’s announcement records that distinction.
The commercial shift was still substantial. Initially, investors could bet on both advances in the company’s models and growth in its consumer product. After key researchers departed, the remaining business faced a more immediate test: could its audience generate durable revenue? Public information does not fully establish how the reported $2.7 billion was allocated among licensing, arrangements associated with the departing talent, and different stakeholders.
Seen this way, August 2024 was more than a change of employer for the founders. Technology and talent had generated value through a major transaction. The consumer platform’s business model still had to prove itself.
03 Beyond Quantifiable Bills Lie Risks That Are Harder to Price
The company’s explanation for its difficulties centers on several pressures: the expense of developing and growing an AI entertainment business, potential liabilities from litigation, and regulatory scrutiny. Together, these pressures affected its cash flow and ability to sustain operations.
The filing placed estimated assets and liabilities each between $50 million and $100 million. Google appeared on the relevant creditor list with approximately $4.15 million owed, alongside law firms, cloud providers, and AI inference providers. Those entries offer a view of the company’s operating obligations, but the listed amounts should not be treated as its entire liability exposure.
Harder to assess are legal claims whose value has not been determined. Entries involving Pennsylvania, Kentucky, and an unnamed minor were identified as contingent, unliquidated, and disputed. An empty amount does not mean an absence of risk. It means that liability, resolution, and the eventual financial consequences remain uncertain.
The disputes date back at least to October 2024, when a Florida mother sued Character.AI following the death of her 14-year-old son, naming Google as a defendant as well. Further actions followed from families and state governments. The company progressively restricted its product, including ending open-ended chats for minors from November 2025. In January 2026, Character.AI and Google reached agreements in principle to settle several family lawsuits, although the full terms were not publicly disclosed.
A potential buyer must evaluate not only existing assets but also possible future obligations. Server bills can be checked line by line; litigation exposure is harder to resolve in a spreadsheet. It can shape what a buyer is willing to acquire, how much it will pay, and which obligations it seeks to exclude from the transaction.
04 Disney Has Put a Price on Part of the Business
Disney is acting as the stalking-horse bidder: the party that establishes an initial set of terms against which competing offers can be evaluated. Its proposal provides a benchmark, but remains subject to court approval and potentially higher or better offers. The $20 million figure therefore cannot be read as the value of Character.AI as a whole.
The disclosed proposal focuses on a non-exclusive, perpetual license to certain intellectual property, certain legal claims, and transition services. “Non-exclusive” does not mean exclusive ownership of the technology. Nor should “claims” automatically be understood as ordinary receivables; their scope depends on the transaction documents. Most importantly, the proposal does not itself settle who will operate the entire platform over the long term.
Disney’s relationship with Character.AI had already evolved considerably. The platform previously hosted numerous characters based on Disney properties. In September 2025, Disney sent a cease-and-desist letter alleging unauthorized use of its characters and raising associated safety concerns. Character.AI subsequently removed the relevant characters.
In September 2026, Disney announced that Character.AI’s then-CEO, Karandeep Anand, would become its newly appointed chief technology officer, effective October 2. It also said that some technical employees were expected to join him. Ten days later, Character.AI filed for bankruptcy, with Colin Adams identified as CEO in the company’s announcement.
The sequence alone does not establish the full arrangements behind these events. It does, however, reveal a similarity between the two episodes: technology and talent attractive to a larger company can find a new home before anyone commits to the continuing operation of the original consumer product.
05 What Matters Most to Users Still Has No Clear Destination
The October 12 user announcement confirmed that the platform remained operational, subscriptions were still active, and billing was unchanged for the time being. Personal information remained subject to the privacy policy, and users could delete their accounts or export their data. The company said it intended to notify subscribers in advance if changes affected paid features.
Those answers describe the immediate situation without fully explaining what happens after a sale or shutdown. Will users still be able to interact with characters they created? Can years of conversations move to another service? How would remaining subscription periods be handled if the platform closes? Which commitments would a new operator carry forward? Those questions require a more detailed plan.
Publicly reported character counts rose from approximately 2.7 million announced by the company in 2023 to more than 18 million alleged in Pennsylvania’s 2026 complaint. The sources and measurement methods are not fully comparable, but they illustrate the scale of user activity. Behind those characters are identities, speaking styles, sample conversations, and substantial investments of time and creativity.
That material may have commercial value and is subject to various rights and privacy arrangements. But exporting data and preserving an experience are different things. Even with every message saved, a user may be unable to recreate the same character, memory, or style of response on another platform.
For users, preserving their data is therefore only one concern. An equally important question is whether the experience they have spent so much time building can continue.
The Gap Between User Value and Business Sustainability
From a business perspective, a consumer AI company’s value lies in its technology and talent, transferable assets and rights, and the content and usage activity generated by its users. These elements are interconnected, but in an asset sale they may attract different buyers and be valued differently, rather than being acquired as a whole.
The 2024 Google transaction reflected the value of technology and certain key personnel through licensing and employee moves. The current bankruptcy sale seeks buyers for particular assets and rights. Users’ accumulated content and interactions may also have commercial value, but preserving those experiences requires ongoing spending on computing, product maintenance, and safety, together with the associated business risks.
The absence of users’ time as a separate balance-sheet asset does not mean users lack commercial value. Equally, time itself is not an accounting expense. More informative measures are the recurring revenue generated by usage and the ongoing expenditure required to support it. If the two remain out of balance, even strong engagement may be insufficient to sustain an independent business.
Character.AI’s experience demonstrates substantial demand for interaction with AI characters, while also showing the gap between demand and business sustainability. A platform must convert engagement into sufficiently reliable revenue, manage service costs, and meet the legal and regulatory responsibilities associated with its product.
The company expects the sale process to conclude in late November or early December 2026. The key development to watch is whether an actionable proposal emerges that supports continued operation—and how it addresses users’ characters, conversations, and subscription rights. The company has said the platform will be wound down if no such proposal is available.
For now, the $20 million proposal provides a benchmark for certain licenses and rights. Whether the platform continues operating and users retain what they have built will depend on the bidding process and the final transaction terms.