AI as Star Witness: How a Buyer’s AI Conversations Sank Its Earnout Avoidance Strategy
On March 16, 2026, the Delaware Court of Chancery issued a significant post-trial opinion in Fortis Advisors, LLC v. Krafton, Inc.1 The case arose from Krafton, Inc.’s (the “Buyer”) acquisition of Unknown Worlds Entertainment (the “Target”), and the Buyer’s subsequent attempt to engineer its way out of a $250 million earnout obligation. The court found that the Buyer breached the parties’ Equity Purchase Agreement (the “EPA”) by terminating the Target’s three key executives without (contractually defined) “Cause” and by seizing operational control of the studio. In a sweeping remedial order, the court reinstated the Target’s CEO to his position with full operational authority, enjoined the Buyer from using the studio’s own board to circumvent that authority, restored the Target’s access to its publishing platform, and equitably extended the earnout measurement period by 258 days. Perhaps most remarkably, the court quoted at length from responses generated by a popular AI platform in response to queries made by the Buyer’s CEO about how to take control of the Target’s operations.
The decision carries important implications for acquirers, sellers, and their counsel in M&A transactions, particularly with respect to the evidentiary risks of using AI tools for legal advice and strategy, for-cause termination rights, the negotiation and enforcement of earnout provisions, and the significance of specific performance provisions in M&A agreements.
Background
In October 2021, the Buyer, a South Korean gaming conglomerate, acquired the Target, which created the Subnautica gaming franchise. Subnautica is an underwater survival adventure game that became a cultural and commercial phenomenon, selling over 17.5 million copies and exceeding $300 million in gross revenue. The Buyer purchased the Target for $500 million upfront, plus up to $250 million in contingent earnout payments tied to post-closing revenue. To close the deal, the Buyer made a critical promise: the Target’s two co-founders and its CEO (collectively, the “Key Employees”), would retain operational control of the studio in all material respects during the earnout period. This included authority over the product roadmap, launch timing, planning, partnering, budgeting, and employee matters. That control would endure so long as any one of the three remained employed.
In exchange, the Key Employees agreed to run the Target in the ordinary course of business, consistent with its historical customs and practices, and accepted that the Buyer’s consent would be required for certain significant corporate actions. The EPA also priced in the possibility that a Key Employee might leave during the earnout period: each voluntary departure would reduce the revenue credited to the earnout calculation by $1 million per departing employee. The EPA included a narrow definition of “Cause” for termination: it was limited to felonies, intentional fraud or dishonesty, gross misconduct, and wrongful disclosure of trade secrets or confidential information, each requiring a high degree of culpability.
After the acquisition closed, the founders gradually transitioned away from day-to-day game development in ways that the Buyer knew about and, in many instances, actively supported. Both founders voluntarily reduced their salaries from nearly $400,000 to $100,000 to reflect their scaled-back roles; those changes were entered directly into the Buyer’s centralized HR system and reviewed by the Buyer’s finance team. The Target’s CEO, by contrast, continued to shoulder the full weight of studio operations, serving as both CEO and the Target’s primary liaison to an increasingly difficult parent company.
By mid-2025, Subnautica 2 was on track for an early access launch in August, and the Buyer’s own internal projections indicated that a successful launch would generate revenue well above the earnout threshold, producing a projected payout of approximately $191.8 million in the base case and $242.2 million in the best case. The Buyer’s CEO felt that the Buyer had agreed to a “bad deal” and that paying the earnout would make him look like a “pushover.” After the Buyer’s own legal team warned him that a for-cause termination would not eliminate the earnout obligation and would expose the Buyer to legal and reputational risk, the CEO turned to an AI chatbot for help. When the chatbot told him the earnout would be “difficult to cancel,” the CEO complained to colleagues that the EPA was “a contract under which we can only be dragged around.” At the chatbot’s suggestion, the CEO formed an internal task force dubbed “Project X,” whose mandate was to either renegotiate the earnout or execute a full ““takeover” of the Target. The AI chatbot furnished the CEO with a detailed “Response Strategy to a No-Deal’ Scenario,” which included a “pressure and leverage package,” strategic talking points for negotiating with the Key Employees, instructions to lock down the Target’s publishing platform and control publishing rights, directions to prepare systematic materials for legal defense, and a “two handed strategy” combining hardball legal and financial pressure with softer retention incentives. The CEO admitted at trial that he had deleted certain of the relevant logs from the AI platform.
The Buyer followed most of the chatbot’s recommendations. It locked the Target out of the publishing platform, effectively severing the studio’s ability to release Subnautica 2. It posted a message on the websites of the Target and Subnautica, which was drafted overnight and without the studio’s involvement, falsely claiming that the founders were considering an invitation from the Buyer to reengage with the project. When negotiations over the earnout stalled in late June 2025, the Buyer determined to execute its takeover strategy. On July 1, 2025, the Buyer sent termination letters to the Key Employees, citing a single ground for dismissal: their intention to proceed with a premature release of Subnautica 2. The Buyer simultaneously removed the Key Employees from the Target’s board, replacing them with Buyer representatives, and installed a part-time replacement CEO for the Target who had never played a Subnautica game and had no experience overseeing early access title development.
The Court’s Analysis
The Terminations Were Not for “Cause”
Vice Chancellor Will of the Delaware Chancery Court found that none of the Buyer’s proffered justifications for the terminations satisfied the EPA’s definition of “Cause.” Under the EPA’s narrowly negotiated definition of Cause, the only provision at issue was whether the Key Employees had committed an “intentional act of . . . dishonesty.” The court interpreted that phrase strictly, holding that intentional dishonesty requires a conscious objective to deceive, not merely a deliberate act that might be characterized as inaccurate or imprudent.
At trial, the Buyer had abandoned its original stated basis for termination (that the Key Employees intended to prematurely release Subnautica 2)and advanced two new theories. The Buyer’s first new theory at trial was that the founders had secretly entered “semi-retirement” and concealed their diminished roles from the Buyer. The court rejected this outright. The record showed that the Buyer had known about, and in many cases actively endorsed, the founders’ evolving roles for over a year before the terminations. The Buyer ’s own internal communications reflected its understanding that one of the founders was not working on Subnautica 2 directly and that his salary had been reduced. The evidence showed no deception, no coverup, and no intent to mislead.
The Buyer’s second new theory was that the Key Employees’ mass downloads of company data constituted terminable offenses for Cause. The court acknowledged that the downloads were “wrongful,” but found they were defensive measures taken in good faith to protect the studio’s work product in the face of the Buyer’s escalating takeover, not acts of intentional dishonesty. Critically, the Key Employees kept the data confidential and returned it promptly after their terminations. As the court put it: “These are not the actions of thieves.”
The court also rejected these new theories by applying the “mend-the-hold” doctrine, which barred the Buyer from relying on justifications it had known about before the terminations but chose not to cite in its termination notices, and the “after-acquired evidence” doctrine, which would have required the Buyer to prove that the later-discovered conduct was so severe that it would have independently resulted in termination if known at the time of discharge.
Because neither of the Buyer’s new theories established the conscious objective to deceive required by the EPA, the court held that the Buyer lacked Cause to terminate the Key Employees.
The Buyer Wrongfully Seized Operational Control
The court separately held that the Buyer’s seizure of operational control was independently wrongful. The EPA conditioned the Key Employees’ right to operational control on three requirements: (a) operating the Target in the ordinary course of business; (b) not taking certain enumerated actions without the Buyer’s prior consent; and (c) exercising operational control in good faith and in material compliance with applicable law. The Buyer argued that the Key Employees had breached all three. The court disagreed with respect to each.
On the ordinary course covenant, the Buyer argued that the Target could not operate “consistent with past custom and practice” without its founders personally leading game development—i.e.,that the founders were the essence of the studio’s “magical pair” creative model. The court rejected this framing. It held that an ordinary course covenant is assessed at the company level, not the individual employee level. Before the acquisition, the founders had already demonstrated a practice of delegating primary development duties to other team members, most notably on Subnautica: Below Zero, which the founders did not lead. After the acquisition, the studio continued to design, develop, and prepare to publish games using its established community-focused, early access development model. The Buyer could not manufacture a covenant breach by pointing to changes in individual executives’ job duties when the company’s fundamental business operations remained unchanged.
On the consent-required actions, the Buyer advanced two theories. First, it argued that the founders’ voluntary role and salary reductions triggered a “constructive termination” under their employment agreements, which required the Buyer’s advance consent under the EPA. The court found this argument commercially unreasonable: the founders could not create a constructive termination claim against themselves by voluntarily reducing their own salaries or modifying their own duties. Second, the Buyer argued that elevating game developers to project-level leadership constituted the hiring of “executive-level” employees requiring the Buyer’s consent. The court disagreed and found that these were routine personnel decisions squarely within the Key Employees’ retained operational authority, not the hiring of corporate executives.
On the good faith obligation, the Buyer argued that the Target’s CEO had encouraged the founders to “stay on the books” in reduced roles specifically to preserve the earnout and that this constituted bad faith. The court was not persuaded. The EPA did not obligate the founders to remain in their exact pre-acquisition roles indefinitely; it merely provided that operational control would persist so long as any one Key Employee remained employed. If both founders had departed, the only consequence would have been a $1 million reduction per departing founder in the revenue counted toward the earnout threshold—a modest adjustment in the context of a potential $250 million payout. In the court’s view, it was entirely reasonable for the Target’s CEO to encourage the founders to transition into reduced advisory roles rather than leave outright. Openly facilitating that transition was not bad faith.
Because the Buyer failed to prove that the Key Employees breached any of the conditions imposed upon their post-acquisition operational control, the court held that the Buyer breached the EPA when it removed their operational control.
The Remedy
Having found that the Buyer breached the EPA on two independent grounds (wrongful termination and wrongful seizure of operational control), the court awarded specific performance, finding that the Key Employees’ right to operational control was a crucial, bargained-for protection that money damages alone could not adequately replace. The EPA itself contained a specific performance clause, and the court held that the Buyer had offered no persuasive or case-specific reason to disregard it. The loss of creative control over a unique asset (the ability to guide the development and launch of Subnautica 2) constituted irreparable harm.
The court crafted a tailored remedy rather than a blanket restoration of the pre-termination status quo. Although the Target’s former CEO was reinstated with full operational authority, the founders were not restored to their pre-termination roles. The court found that restoring the Target’s former CEO alone was sufficient to vindicate the Key Employees’ contractual rights since they would still qualify for the earnout and the founders could still be engaged in an advisory capacity. The court declared that a previous resolution by the Buyer’s board, which had purported to prohibit the release of Subnautica 2 absent board approval, was ineffective to the extent it infringed on the Target CEO’s operational authority. The Buyer was enjoined from using the Target’s board, or any other corporate mechanism, to circumvent the Target CEO’s decision-making authority over the game’s early access launch. The Buyer was also ordered to immediately restore the Target CEO’s access to the Target’s publishing platform.
To ensure the specific performance remedy was not rendered illusory by the passage of time, the court also equitably extended the earnout measurement period by 258 days, the exact duration of the Target CEO’s ouster measured from his July 1, 2025 termination to the date of the opinion. Whether the Buyer’s actions actually impaired the earnout, and what money damages (if any) are owed, remain reserved for a forthcoming second phase of the bifurcated litigation.
Key Takeaways
AI communications may be discoverable and provide unique insight into a party’s motives. Perhaps the most striking feature of the opinion is that the court quoted at length from the Buyer CEO’s conversations with the AI chatbot and expressly relied on those exchanges to establish the pretextual nature of the Buyer’s conduct. The chatbot’s recommendations were treated as a window into the Buyer’s strategic intent. The court also noted that the Buyer’s CEO admitted at trial to deleting specific, relevant logs from the AI platform—a fact that may factor prominently in the second phase of the litigation, where money damages and potential earnout impairment are at issue. The opinion does not establish new legal standards governing AI use, but it does carry a clear warning: communications with AI platforms may not be privileged, may be subject to discovery, and may be used against the party that generated them. Deleting them may compound the problem. Acquirers using AI tools for post-closing strategy or any deal-related purpose should treat those communications with the same discipline they would apply to other non-privileged communications, including applying an appropriate retention policy.
The evidentiary and protected status (or lack thereof) of AI-generated communications is an area of law that is rapidly developing. A recent Mayer Brown Legal Update (M&A Discovery in the AI Era: Generative AI Communications and Outputs May Become Litigation Ammunition) examines in depth recent judicial opinions on the application of privilege to AI use. The landscape is also evolving legislatively: a proposed New York bill introduced in March 2026 would bar AI chatbots from posing as lawyers and give deceived users the right to sue for damages. The Krafton case provides a vivid, real-world illustration of why practitioners and dealmakers should be paying close attention to these developments.
Be careful what you put in your termination notice because you will likely be bound by it. The Krafton case is a cautionary tale about what happens when, in the middle of litigation, an acquirer tries to substitute its stated basis for termination with new grounds. The court applied the mend-the-hold doctrine to bar the Buyer from relying on the founders’ role changes as a justification for termination when the Buyer had known about those changes long before the terminations and chose not to cite them in its termination notices. For the data downloads, which the Buyer did not fully discover until after the terminations, the court applied the after-acquired evidence doctrine, which required the Buyer to prove that the downloads independently would have resulted in termination. The Buyer failed that test as well. More broadly, the court interpreted the EPA’s “intentional act of dishonesty” standard as requiring a conscious objective to deceive, not merely deliberate conduct that could be characterized as improper. Acquirers should carefully consider whether their for-cause termination definitions are broad enough to cover the full range of conduct they may realistically encounter post-closing and should understand that the grounds cited in the termination notice are likely the grounds they will be stuck with at trial. Sellers should continue to push for narrow, specifically negotiated definitions of cause.
A breach during the earnout period can cost more than the breach itself. The Krafton decision demonstrates that Delaware courts may use their broad equitable powers to extend earnout measurement periods when a buyer’s breach has deprived the sellers of the time and operational runway they were promised. In Krafton, the court extended the base earnout measurement period by 258 days—the exact duration of the improper ouster of the Target’s CEO. The extension was calibrated to ensure that the specific performance remedy was not illusory: sellers need a genuine opportunity to achieve earnout targets, and a court-ordered reinstatement without time to generate the required revenue accomplished little. For acquirers contemplating post-closing interventions, the cost of a breach is not merely the price of lost litigation but is also the cost of additional time on the earnout clock.
Specific performance clauses carry real weight in Delaware. Krafton underscores that Delaware courts will enforce a specific performance clause unless the breaching party offers a “persuasive and case-specific” reason to override it. The court not only reinstated the Target’s CEO but also issued broad injunctive relief preventing the Buyer from using the Target’s board of directors to circumvent the Target CEO’s operational authority. The takeaway for sellers is that specific performance clauses are one of the most valuable protections in an earnout-heavy deal structure and should be negotiated robustly. The takeaway for acquirers is that the equitable remedies Delaware courts are willing to craft may extend well beyond simple reinstatement; they also encompass injunctions, earnout period extensions, and the nullification of board resolutions, leaving little room to maneuver post-breach.
1 ___ A.3d. ___, C.A. No. 2025-0805-LWW (Del. Ch. March 16, 2026) (Will, V.C.).
