When Changhan Kim, CEO of South Korean gaming giant Krafton, decided he had agreed to a contract that was too generous, he did not go to his lawyers first. He went to ChatGPT. What followed is now the subject of a Delaware court ruling that has embarrassed one of the world’s most profitable gaming companies and handed a cautionary tale to boardrooms everywhere.
Also read: The “Godfather of AI” quits Google, warns of dangers of misinformation
What transpired
The story begins in 2021, when Krafton, the publisher behind the global phenomenon PUBG: Battlegrounds, acquired Unknown Worlds Entertainment, the US studio behind the beloved underwater survival game Subnautica, for $500 million.
Baked into that deal was an earnout clause: an additional $250 million in performance payments tied to the successful launch of Subnautica 2. The founders of Unknown Worlds, Ted Gill, Charlie Cleveland, and Max McGuire, retained operational control of the studio and could only be removed for legitimate cause. It was a clean arrangement, one that Krafton’s own legal team understood well.
By mid-2025, however, as Subnautica 2 climbed the Steam wishlist charts and the $250 million earnout began to look increasingly real, Kim had developed a different view of the contract. He considered it, according to court documents, a deal he had been too soft on. His goal was to find a way out. When his head of corporate development, Maria Park, advised him bluntly that firing the founders without cause would not void the earnout and would very likely trigger a lawsuit, Kim turned not to outside counsel or his internal legal team. He turned to ChatGPT.
Can’t spell fail without AI
The chatbot’s initial response was not encouraging. It told Kim the earnout would be difficult to cancel.
Dissatisfied, Kim pushed further. ChatGPT eventually suggested he form an internal task force to either negotiate a revised deal with Unknown Worlds or execute a full corporate takeover of the studio. Kim acted on both. The task force was codenamed Project X. Its mandate was to buy time, restrict Unknown Worlds’ operational autonomy, and if necessary, wrest control of Subnautica 2 entirely.
In July 2025, Krafton fired Ted Gill, Cleveland, and McGuire. The stated reasons were launch delays and unauthorised side projects. The court would later find those reasons to be, in its words, clearly pretextual.
What Kim did not account for was that his ChatGPT conversations, which he subsequently deleted, were recoverable and discoverable. They became central evidence in the lawsuit that Gill filed in the Delaware Court of Chancery. The deleted logs showed the CEO consulting an AI chatbot to contrive a corporate takeover strategy, a phrase the judge, Delaware Vice Chancellor Lori Will, used verbatim in her March 16, 2026 ruling.
The ruling went almost entirely against Krafton. Vice Chancellor Will found that the company had breached its Equity Purchase Agreement by terminating the founders without valid cause. She ordered Ted Gill reinstated immediately as CEO of Unknown Worlds, with full operational authority over Subnautica 2, including control over its early access launch on Steam. She extended the earnout measurement period by 258 days, pushing the deadline to September 15, 2026, with a possible further extension to March 2027. A second phase of proceedings will determine the precise damages owed.
The ruling landed with particular force because Kim had acted against the explicit advice of his own legal team. When Park warned him the plan was likely to fail and produce litigation, Kim chose the AI that told him what he wanted to hear. ChatGPT produced a detailed, plausible-sounding strategy. It had no liability, no professional obligations, no knowledge of Delaware contract law, and no capacity to tell Kim that his underlying goal was itself legally indefensible.
The irony compounds when you consider Krafton’s position. The company reported record revenue of over $2.26 billion in 2025. It was not a business under financial pressure. The decision to pursue Project X was not one of necessity but of opportunism, and the AI-generated playbook it followed made the opportunism legible to a judge in ways that a more cautious, lawyer-led approach might not have.
The lesson here is not that artificial intelligence has no place in corporate decision-making. It clearly does, as a drafting tool, a research aid, a sounding board for early-stage thinking. The lesson is that a CEO consulting a chatbot instead of qualified legal counsel, on a $250 million contractual obligation, and then acting on its output, is not a technology failure. It is a judgment failure. ChatGPT did not breach the Equity Purchase Agreement. Changhan Kim did. The chatbot simply gave him a roadmap and he chose to follow it, all the way to a Delaware courtroom.
