LM

Buyer Bets on ChatGPT, Suffers Embarrassing Loss

Leonard J. French|

ANALYSIS | MERGERS & ACQUISITIONS / DELAWARE CORPORATE LAW

Fortis Advisors v. Krafton is a case about a $250 million earnout, a video game sequel held hostage, and what Delaware courts do when a sophisticated buyer decides the contract it signed no longer suits it.

By Leonard French | March 16, 2026

There is a version of this story where Krafton, Inc. pays $250 million to the founders of Unknown Worlds Entertainment, everyone walks away having honored their agreement, the rest of us are enjoying Subnautica 2 in early access — and none of us ever hear about it. That is not what happened.

What happened instead produced one of the more striking opinions to come out of the Delaware Court of Chancery in recent years — a detailed, sometimes withering account of how a South Korean gaming conglomerate tried to engineer its way out of a contractual obligation it found inconvenient, and how thoroughly that effort collapsed under scrutiny.

Vice Chancellor Morgan Will, for the Court, issued the ruling today. It is worth reading in full. For those who cannot, what follows is an attempt to explain what the case was about, what the court decided, and why it matters.

The Setup

Unknown Worlds Entertainment makes Subnautica — an underwater survival game with an unusually devoted following. The studio was founded by Charlie Cleveland, a developer whose signature method involves releasing games in unfinished “early access” form, gathering player feedback, and refining the product over years in close collaboration with the community. It sounds counterintuitive. It has worked, repeatedly, to remarkable effect.

In 2021, Krafton — the South Korean publisher behind PUBG: Battlegrounds — acquired Unknown Worlds for $500 million upfront. That figure alone reflects the significant value of the deal. But the deal also included an earnout: a promise to pay up to $250 million more, depending on the studio’s revenue through December 31, 2025. If Unknown Worlds’ new title performed well enough to cross a revenue threshold, Krafton would owe an additional $3.12 for every dollar beyond it, up to the cap.

Earnouts are common in acquisitions where the parties disagree on what the business is worth. The seller believes in its own future. The buyer is skeptical, or at least cautious. The earnout bridges the gap: if you’re right, you get paid. But earnouts create a structural problem that every sophisticated seller understands. Once the deal closes, the buyer runs the company. A buyer who grows reluctant to pay a large earnout can delay product launches, starve marketing budgets, or otherwise manage operations in ways that depress revenue — while technically staying within the letter of the agreement.

The Unknown Worlds founders were not naive about this. As a condition of the sale, they negotiated an explicit contractual guarantee: operational control of the studio would remain with the three “Key Employees” — Cleveland, co-founder Max McGuire, and CEO Ted Gill — for the duration of the earnout period. Krafton could remove them only for narrowly defined “Cause.” So long as any one of the three remained employed, the others’ authority over product development, launch timing, and business operations was protected.

This clause was not boilerplate. It was the cornerstone of the deal from the sellers’ perspective, and Krafton’s willingness to agree to it was part of what made Krafton the chosen buyer over thirty other interested parties.

The Earnout Becomes a Problem

For a year or two after the acquisition, things went reasonably well. Then came Moonbreaker, an ambitious digital miniatures game the founders had been developing. It launched in early access in September 2022 to decent reviews and then, quietly, died. Players did not show up in the numbers needed to sustain it. Cleveland, who had poured years of creative energy into the project, was devastated. By late 2023 he was burned out enough that he told his colleagues at a company retreat that he had “hit a wall.” Krafton’s own representatives were present for that conversation.

Cleveland stepped back from daily game development. He began spending most of his limited working hours on a film adaptation of Subnautica — a transmedia strategy that Krafton itself had encouraged as a way to grow the franchise’s audience. McGuire, meanwhile, found his technical role had grown beyond his interests as the studio scaled; he pivoted toward research on how video games affect children with autism. Gill kept running the business.

All of this was disclosed to Krafton. The title changes formalizing Cleveland and McGuire’s reduced roles were communicated directly to Krafton’s head of corporate development and acknowledged by the CEO. Cleveland’s Hollywood work was documented publicly on his own social media, in podcasts linked directly to Krafton’s leadership, and in direct correspondence. The salary reductions — both founders went from roughly $400,000 to $100,000 annually — were entered into Krafton’s own HR system by Gill and reviewed by Krafton’s finance team.

None of that mattered once Krafton saw the earnout projections.

In May 2025, with Subnautica 2 approaching its planned August early access launch, Krafton’s finance team modeled out the scenarios. A successful launch would likely generate between $191 million and $242 million in earnout payments. CEO Changhan Kim, who had personally championed the acquisition, reacted with something close to alarm. Internal messages show him describing the EPA as a “bad deal” and expressing frustration at feeling like a “pushover.” He told Cleveland in a meeting that paying the earnout could reduce the studio’s book value — something he would personally be held accountable for.

(Quick sidebar: if the EPA, which was part of the purchase agreement, reduces the book value of your business that you just purchased, then I argue you are using the wrong book value. You are using a value adjusted for what you think the company is worth without a key liability. Some people might consider omitting a key liability from your book value to be some kind of misrepresentation or fraud.)

His legal team told him, accurately, that firing the Key Employees would not automatically eliminate the earnout. So Kim consulted ChatGPT.

Project X

This is the part of the opinion that will be quoted in law school books for years.

ChatGPT confirmed that the earnout would be difficult to cancel. Kim complained to his head of corporate development that the EPA was “a contract under which we can only be dragged around.” He then used the AI to generate a “Response Strategy to a No-Deal Scenario” — a document that included a “pressure and leverage package,” suggestions for “securing control points” by locking down Steam publishing rights, recommendations for building “systematic materials for legal defense,” and a proposed public messaging strategy to undermine the studio’s leadership with its own fanbase.

Krafton then followed most of it. An internal task force, dubbed “Project X,” was formed. Krafton seized control of the Steam account through which Unknown Worlds published its games, cutting off the studio’s practical ability to release Subnautica 2. It posted a message on the Unknown Worlds website — without the team’s knowledge — stating, falsely, that Cleveland and McGuire were “considering” an invitation to return to lead the game they had never abandoned. It began compiling “supporting materials” consisting of Cleveland’s social media posts about his film work.

When negotiations over a buyout of the earnout stalled, Krafton’s head of strategy told Kim: “It might actually be easier to just do a takeover.” Kim replied: “Set a date.”

On July 1, 2025, termination letters went out to all three Key Employees. The stated reason: their “intention to proceed with a premature release of Subnautica 2.”

What “Cause” Means, and Why It Mattered

Here is the legal heart of the case. Krafton had the contractual right to fire the Key Employees, but only “for Cause” as that term was defined in the EPA. The definition listed four specific grounds: felony conviction, intentional wrongful disclosure of trade secrets, willful acts constituting gross misconduct, and — the one Krafton ultimately relied upon — “an intentional act of fraud or dishonesty.”

The Vice Chancellor’s analysis of that phrase is careful and worth understanding. “Dishonesty” is not simply being wrong, or inaccurate, or even unauthorized. It requires a deceptive state of mind — a disposition to deceive. The word “intentional” elevates that further: it cannot be mere surplusage, so it must mean something beyond dishonesty alone. Reading the provision in context, alongside the other Cause definitions involving felonies and gross misconduct, the court concluded that “intentional dishonesty” requires a conscious objective to deceive. Not sloppiness. Not a policy violation. A deliberate aim to mislead.

Against that standard, Krafton’s two arguments for Cause failed completely.

The first — that the founders had secretly slipped into semi-retirement while pretending to run the studio — collapsed against the documentary record. Krafton knew. It knew in February 2024 that Cleveland would not be working on Subnautica 2. It knew by July 2024 that the title changes formalized roles that had already been operating for months. Its own internal communications show its executives understood that “there is not really much change” beyond titles that “more clearly represent the actual roles.” The salary data was in Krafton’s own systems. Park, Krafton’s head of corporate development, had been receiving direct updates throughout. The argument that any of this was concealed from Krafton was not a close call — the court treated it as simply not credible.

The second argument — the data downloads — required more analysis, because the conduct itself was genuinely problematic. In the weeks before their terminations, Cleveland, McGuire, and Gill each downloaded large volumes of company data to personal devices. Cleveland downloaded 72,000 files including proprietary source code. McGuire downloaded nearly 100,000. Gill exported his email and Slack messages. These were not routine backups.

The court found them wrongful. But it also found they were not dishonest. By late June 2025, Krafton had locked the studio out of its publishing platform and was, by its own internal communications, executing a takeover. The Key Employees reasonably feared they were about to lose access to everything they had built. They downloaded the data to protect it, kept it confidential, and returned it when asked. They did not steal it to start a competing company. They did not sell it. The absence of any deceptive purpose was, to the court, dispositive.

The Shifting Justifications

There is a procedural dimension to this case that deserves attention even though it involves legal doctrine rather than dramatic facts. Krafton’s termination notices cited one reason: premature game release. During litigation, Krafton abandoned that ground entirely and substituted two others — the role changes and the data downloads.

Delaware courts have developed two doctrines specifically to deal with this kind of substitution. The mend-the-hold doctrine prevents a party from swapping out its contractual justification once litigation begins and the original grounds start to look weak. If you knew about the conduct before you pulled the trigger and chose not to cite it, you have waived it. The after-acquired evidence doctrine is the companion rule for genuinely new discoveries: misconduct you learn about only after a termination can still justify it, but only if you prove the conduct was severe enough that you would have fired the employee on that basis alone.

Krafton failed on both. The role changes were known well before July 1. The data downloads, which Krafton did not fully learn about until a post-termination IT audit, failed the after-acquired evidence test because they did not independently satisfy the EPA’s Cause definition — and because the court found Krafton’s claim that it would have terminated over a data backup, independent of the earnout dispute, simply not believable given the documentary record.

Courts are not supposed to be naive, and Vice Chancellor Will was not. When an employer is facing a nine-figure contractual liability and goes looking for reasons to fire the people entitled to it, the reasons it finds are examined with appropriate skepticism. The internal messages, the deleted ChatGPT logs, the Project X documents — all of it built a picture of an employer who had decided on a result and worked backward to a justification.

The Remedy

Specific performance — meaning the court orders a party to actually do what it promised, rather than simply pay damages — is available in Delaware when monetary compensation cannot adequately remedy the harm. Here, the loss was not just revenue. It was the loss of creative control over a unique studio during a finite and critical period. You cannot fully compensate that with a check after the fact.

The EPA itself made the case for specific performance explicit: the parties had agreed that a breach would cause irreparable harm and that specific performance would be an appropriate remedy. Delaware courts treat such provisions seriously.

The court’s relief was precise. It did not restore all three Key Employees — Cleveland and McGuire had moved on to genuinely peripheral roles, and their return was unnecessary to vindicate the sellers’ rights under the EPA. What the contract required was that operational control exist so long as any Key Employee remained employed. Restoring Gill alone accomplished that. He was reinstated as CEO with full operational authority, including over the launch of Subnautica 2. The July 1 board resolution purporting to block that launch was declared ineffective. Krafton was enjoined from using the Unknown Worlds board to circumvent Gill’s authority. The Steam platform access must be returned immediately.

Most significantly for the sellers’ economic interests, the earnout Testing Period was extended by 258 days — the exact duration of Gill’s wrongful removal. The base deadline moves from December 31, 2025 to September 15, 2026, with Fortis retaining the contractual right to extend further to March 15, 2027. Without this extension, the specific performance remedy would have been largely symbolic: Gill could have been reinstated into a window that had already closed.

What Have We Learned?

The deeper lesson of Fortis v. Krafton is not about ChatGPT. The AI is a detail, though an unusually vivid one. The lesson is about what earnout protections actually mean when they are drafted carefully and enforced seriously.

Buyers in earnout deals hold structural advantages. They control the business. They control the resources, the launch calendar, the marketing spend. Sellers who understand this try to negotiate protections — operational control clauses, anti-interference provisions, carefully defined Cause standards — that limit the buyer’s ability to game the revenue calculation. Krafton agreed to exactly those protections. It then apparently concluded, when the earnout projections came in, that the contract it had signed was a problem to be solved rather than an obligation to be honored.

Delaware took a different view. The Cause definition the sellers negotiated was narrow by design. “Intentional dishonesty” means something specific, and the court held Krafton to that meaning rather than stretching it to cover conduct that was, at most, a policy violation driven by defensible motives. The mend-the-hold and after-acquired evidence doctrines prevented Krafton from litigating its way to a justification it had never relied on at the time of the termination. And the equitable extension of the earnout window ensured that the remedy was real rather than nominal.

For lawyers drafting deal documents, a few things are worth noting. The Cause definition here was the product of specific negotiation — the opinion says so explicitly. Vague termination standards favor buyers; sellers should insist on specificity and make sure the definition of terminable conduct is tied to serious culpable intent, not mere policy deviation. Role transitions post-closing should be documented, communicated, and contemporaneously acknowledged by the buyer — not because Delaware law requires it, but because the absence of that documentation is where “concealment” arguments find their purchase. And parties should understand that the mend-the-hold doctrine is not a procedural technicality; it is a substantive constraint on how disputes are litigated.

The earnout and damages questions go to Phase Two. Krafton may yet face liability for the revenue lost during the months it controlled the studio. That litigation will be watched closely.

For now, Ted Gill has his job back, Subnautica 2 can ship when he decides it is ready, and a Delaware court has reminded the market that a contract is not a suggestion.


Unknown Worlds head Steve Papoutsis shared the following message with the Subnautica 2 team today:

“Team, We want to take a moment to sincerely thank every member of the Unknown Worlds team and our partners. Over the past nine months, you have worked incredibly hard to bring Subnautica 2 close to the finish line. That effort has resulted in a game we are truly proud of, one that each of us and our partners at KRAFTON unanimously determined is ready for Early Access release in May. Thanks to your dedication and talent, we’ve added more story chapters, built new creatures, and created new biomes along with many other features. With this significant progress, we have passed KRAFTON’s milestone review last week and are now ready to start our open development journey alongside our community. We look forward to working with Ted Gill to support a smooth transition and work toward a successful launch.”

Krafton itself appears to be on board, per a statement shared with IGN:

“We can confirm that the internal message shared with the Unknown Worlds team is accurate. The team has made meaningful progress following the milestone review approved earlier this month, and has been working toward an Early Access release in May. While we respectfully disagree with this week’s court ruling and intend to explore all legal avenues, our goal is to minimize disruption to the team and to the release plan. We remain committed to an open development approach, working closely with the community as development continues, and are focused on delivering the best possible experience to players.”

I think this is a great outcome for both parties, including Krafton. It seems obvious to me that their CEO was not operating in the best interests of shareholders — something many of us gamers suspected when we learned that Ted Gill and others confirmed Subnautica 2 was ready for early access. I suspected from the start that the $250 million payout had far more to do with it than Krafton was letting on.

So now gamers get the game, released the way its developers intended; its creators get their potential earned-value-payout if the game does well enough; and Krafton gets the ship righted in time to maybe save the franchise.


I’m Leonard French, your favorite copyright attorney. Let me know your thoughts in the comments below.

Lawful Masses is community supported. Special thanks to Eevi, Ugly Grill, TechTechPotato, The Blood Soaked Survivors, and Kyle Siefring for their top-tier support.

This article is for informational purposes only and does not constitute legal advice. The opinion discussed is Fortis Advisors, LLC v. Krafton, Inc., C.A. No. 2025-0805-LWW (Del. Ch. Mar. 16, 2026).

#Delaware corporate law#mergers and acquisitions#earnout#Krafton#Subnautica#Unknown Worlds#contract law#specific performance