Two numbers, two different kinds of commitment
The number attached to this deal since 12 October has been $20 million: Disney's stalking-horse bid for some of Character.AI's assets, centred on a non-exclusive, perpetual licence to certain intellectual property. A stalking-horse bid is an opening floor in a bankruptcy auction, not a completed purchase, and rivals can still bid above it.
A legal filing in the Chapter 11 case, reported by TechCrunch on Sunday after Bloomberg spotted it, adds the second number. Disney has agreed to make employment offers to at least 60 Character.AI employees — the majority of the workforce — at an estimated $60 million in annual compensation.
Keep the kinds straight, because the comparison invites a mistake. The $20 million is a proposed price for specified assets and rights. The $60 million is an estimate of what the offered jobs would cost per year in pay. The second figure being three times the first does not mean Disney values the team at three times the technology; a recurring pay estimate and a one-off asset price are not the same kind of quantity. What the filing does establish is scale: recruiting these employees is a substantial financial commitment in its own right, separate from the asset bid. For the bankruptcy background — the filing itself, the missing buyer for the platform, and the wind-down warning — see our 13 October report: https://alfw.ai/article/character-ai-bankruptcy-disney-20m/
The per-person arithmetic, with its limits
Dividing the reported $60 million estimate by the minimum 60 offers gives roughly $1 million per person per year. Treat that as illustrative arithmetic (ours), not a filing figure and not a verified average. The 60 is a floor: if the estimate covers more than 60 people, the per-person figure is lower. The reporting also does not break the estimate down between salary, equity and other compensation, and it does not identify the seniority or job mix of the people receiving offers. What can be said is narrower: the scale of the commitment suggests that hiring Character.AI's staff is a substantial part of Disney's plan. The composition of that commitment is not public.
Most staff accepted. The platform is still unsold.
The filing says the "vast majority" of the employees who received Disney offers accepted them. Where most of the team is headed now looks clear. When they start, and on what terms, is not in the reporting: accepting an offer is not the same as having started work.
Nothing in the filing changes the platform's position. Disney's bid is for specified assets and intellectual-property rights. It is not an offer to buy the company, and it carries no commitment to keep the Character.AI app running. As we reported on 13 October, the company has said the platform will be wound down when the sale process ends if no bid emerges to keep it operating as a going concern. The staff question has an answer taking shape. The user question — what happens to the service, its characters and its chat histories — is untouched by the hiring news.
The layoff contrast, stated carefully
The offers also arrive shortly after reported layoffs at Disney affecting several hundred employees, including technology workers, according to Business Insider reporting summarised by Benzinga. Set side by side, the two moves raise an obvious question about Disney's technology priorities: reducing some technology roles while offering an outside AI team jobs at an estimated $60 million a year in compensation. But the reporting does not establish that the layoffs were made to finance or make room for the Character.AI hires, and the two should not be read as a single staffing decision without that evidence.
What the filing says about why Character.AI failed
The filing also describes, in the company's own case, the pressures that brought a heavily used product to a bankruptcy court: high costs, and "potential litigation-related liabilities and regulatory engagement," despite a "large and engaged user base." Read at that level, it underlines a distinction worth keeping: attracting users and sustaining a viable service are different achievements.
What the disclosure does not do is isolate a single decisive cause. It does not separate inference costs from subscription economics, or either from legal exposure. The company's legal and child-safety problems are part of the record of this bankruptcy in their own right — the lawsuits and investigations were covered in our 13 October report — and they should not be compressed into a story about the price of compute.
An echo of Google's deal, not a repeat
This is the second time a very large company has taken Character.AI's technology and people while leaving the platform behind. In 2024, Alphabet disclosed in its Form 10-K that it paid $2.7 billion in cash for a non-exclusive licence to Character.AI's then-current model technology, alongside hiring some of its employees, including its founders. Disney's arrangement echoes that structure at a very different scale and in very different circumstances: a proposed licence and asset purchase out of bankruptcy, plus employment offers to most of the remaining staff. The resemblance is structural. It does not make the two deals the same valuation of the same business, and neither deal priced the consumer platform itself. In both cases, that platform was the part left to find its own buyer.
Sources
- [1] TechCrunch, Anthony Ha, "Disney agrees to hire the majority of Character.AI's workforce", 18 October 2026 — the filing's offer numbers, the estimated $60M annual compensation, the vast majority accepted, the stalking-horse status, the filing wording on costs, liabilities and regulation, and Anand as Disney's first CTO; filing first spotted by Bloomberg, case documents via StrettoRead source
- [2] alfw.ai, "Character.AI Had Millions of Users. It Filed for Bankruptcy Anyway.", 13 October 2026 — our earlier coverage: the $20M licence structure, stalking-horse mechanics, the wind-down warning and the sale timingRead source
- [3] Benzinga, summarising Business Insider's reporting — used only for the Disney layoffs contextRead source