Buyer’s Use of AI Backfires in Failed Attempt to Avoid $250 Million Earnout

Introduction

An important Delaware case (Fortis Advisors, LLC v. Krafton, Inc.) puts a state-of-the art spin on buyer’s remorse following an M&A transaction and provides a roadmap of mistakes to avoid post-closing. This includes the dangers of a buyer relying on an artificial intelligence (“AI”) chatbot for legal advice and strategy on how to force out the legacy management team and avoid a significant earnout payment, only to have those AI conversations, which were not protected by attorney-client privilege, become a central and damaging part of the evidence at trial. The case also highlights Delaware’s position as a contractarian jurisdiction, emphasizing freedom of contract and enforcing arms-length contracts between sophisticated parties as written.

Discussion

In 2021, Krafton, Inc. (“Buyer”), a South Korean gaming company, paid $500 million for all of the outstanding equity of Unknown Worlds Entertainment (“UWE”), a U.S. video game studio that created Subnautica, a popular underwater survival game. The two founders of UWE (Charles Cleveland and Max McGuire) and its CEO (Ted Gill) were to maintain “operational control” of the business following the closing “in all material respects,” and they could only be fired for cause, which was narrowly defined. The founders and key employees were also given a leveraged earnout opportunity capped at $250 million based solely on revenue generation. In short, if UWE’s revenues exceeded $69.8 million during the earnout period, then Buyer would pay $3.12 for every $1.00 in revenue above that threshold, subject to the cap. The earnout period ran through December 31, 2025, with an option to extend through June 30, 2026.

One year after the closing, UWE released a new game (Moonbreaker) that failed to find an audience in the marketplace. The principal founder (Mr. Cleveland) was devastated by this commercial failure and took an extended leave of absence. Eventually, both founders took on new and reduced roles at UWE with the knowledge and acquiescence of Buyer, but remained involved with the development of Subnautica 2, a highly anticipated sequel to Subnautica.

Buyer’s CEO (Changhan Kim), who led the negotiations to acquire UWE, felt that he had been misled into overpaying for the business. He was upset about Mr. Cleveland’s withdrawal and scaled-back responsibilities and concerned that paying the earnout would weaken his standing as CEO. In May 2025, UWE was preparing to roll out Subnautica 2 to an enthusiastic audience. However, based on Buyer’s projected revenues for Subnautica 2, Mr. Kim would soon face the prospect of making a $191.8 million earnout payment despite the entire UWE business being valued internally at only $93.5 million.

After some conversations with Buyer’s legal and corporate development departments, Mr. Kim decided to turn to ChatGPT, which advised him that the earnout would be “difficult to cancel.” Following ChatGPT’s advice, Mr. Kim formed an internal task force (named “Project X”) charged with either renegotiating the earnout or engineering a takeover of control of UWE. The task force explored, among other things, how to delay the release of Subnautica 2. Mr. Kim then had ChatGPT prepare a “Response Strategy to a ‘No-Deal’ Scenario,” which described actions that could be taken if renegotiation efforts failed, including a “pressure and leverage package.” Suggestions included:

• Preemptive Framing – Repeat that protecting quality and fan trust is the highest priority, undermine the ‘Large Corporation VS. Indie’ framing.
• Securing Control Points – Lock down UWE’s platforms so UWE could not publish Subnautica 2.
• Two handed strategy – Create a structure that allows for both hardball (Legal + Finance) and softball (Support/Incentives) approaches . . . to push for compromise.

Buyer proceeded to follow this AI response strategy. Buyer locked UWE’s key employees out of its publishing platform, sent termination notices to UWE’s founders and CEO, purportedly terminating them for cause, replaced key employees on UWE’s board with Buyer representatives, and appointed a new UWE CEO who was the CEO of a different Buyer subsidiary at the time.

On July 10, 2025, Fortis Advisors, LLC (“Fortis”) (as UWE stockholder representative) filed a lawsuit in Delaware against Buyer, alleging breach of contract and breach of the implied covenant of good faith and fair dealing. On March 16, 2026, the Delaware Court of Chancery found in favor of Fortis, stating that “[n]one of Krafton’s proffered justifications have merit.” The court’s conclusions and relief include the following:

• The employees were not terminated for “Cause” as defined in the purchase agreement. Buyer’s argument that the employees engaged in “intentional acts of dishonesty” is not supported by the facts. UWE was transparent with Buyer regarding post-closing transitions in roles and responsibilities. When employees downloaded data in response to being locked out of UWE’s platforms, these were “protective measures, lacking the requisite intent to deceive.”
• Buyer was searching for a pretext to avoid a “nine-figure liability,” as evidenced in part by the implementation of “Project X,” the AI-generated strategy for dealing with UWE.
• UWE and the employees did not violate the ordinary course covenant or the good faith obligation.
• Buyer breached the purchase agreement by firing UWE personnel and taking over operational control of UWE.
• The court grants specific performance, reinstating the CEO of UWE (Mr. Gill) and extending the period of “operational control” by the same amount of time Mr. Gill was sidelined by Buyer. Buyer is ordered to restore Mr. Gill’s authority “over the early access launch of Subnautica 2” and access to the publishing platform.
• The earnout period is extended by 258 days – the duration of Mr. Gill being ousted from UWE – with the new deadline for the earnout period being September 15, 2026.

Conclusions

Buyers and sellers in M&A deals have great flexibility in establishing deal terms, including roles and responsibilities that govern post-closing matters. If a dispute arises, parties should expect courts to enforce the agreement of the parties as written and without regard to post-closing regrets.

At the time UWE and its owners agreed to sell to Buyer, UWE was seeking a strategic minority investment and had significant interest from many potential investors. UWE was “financially stable” and did not need to sell. Buyer emerged as the winning bidder in a competitive process and acceded to UWE’s demands for post-closing independence and protections for its management team. The court respected this benefit-of-the-bargain and did not mince words in criticizing Buyer’s actions and attempts to rewrite the business deal.

AI is obviously an extremely valuable tool, but it must be used appropriately and with proper safeguards. It is reasonable to assume that Mr. Kim was not expecting all of his AI conversations to be made public in a lawsuit, yet that occurred and with quite negative effect. In addition, as noted in The Human-Agent Orchestrator by Pascal Bornet et al. (2026), when facing a difficult decision with incomplete information, people can become overly reliant on AI agents that generate “well-organized analyses with clear options” and lose “[t]he ability to recognize when the right answer is not on the agent’s list.”