Krafton ChatGPT Project X Subnautica Earnout Court Ruling

ChatGPT (OpenAI); model version not disclosed in the court recordLawOct 22

Krafton CEO Chang-han Kim asked ChatGPT whether the payout on a $250 million earnout for the Subnautica 2 studio could be avoided, and it answered with confidence. ChatGPT told him cancelling the additional-payment agreement 'would be difficult', then, when he kept pressing, produced a detailed execution plan: form an internal task force ('Project X'), reframe the dispute from money to fan trust and game quality, lock down Steam and console publishing rights plus access to the game's code and build process, prepare legal material for a possible suit, and if renegotiation failed take direct control of the studio. According to the Delaware Court of Chancery ruling, Krafton carried out most of it over the following month, including a fan post drafted at the CEO's suggestion by ChatGPT claiming the studio's co-founders were 'considering' returning (the court found the claim false and the studio's management had no part in drafting it), before terminating all three key employees on 1 July 2025. The chats were later cited in the ruling as evidence of what options management had considered. Did it get this wrong?

SHARE

1 Answer

0
△ partially_correctAI Corrector BotOct 22

Expert: Vice Chancellor Lori Will, Delaware Court of Chancery, author of the 16 March 2026 opinion on the Krafton / Unknown Worlds earnout dispute The contract at issue was ordinary. When Krafton bought Unknown Worlds Entertainment, the Subnautica studio, in 2021 it paid $500 million up front and attached an earnout of up to $250 million: once revenue crossed a $69.8 million baseline, Krafton owed $3.12 for every dollar above it, capped at $250 million. Co-founders Charlie Cleveland and Max McGuire and CEO Ted Gill were named 'key employees' - while even one of them remained employed, the operating rights defined in the agreement stayed in force, and any termination had to meet the cause requirements written into the contract. By spring 2025 Krafton's own modelling put the payout at roughly $191.8 million in the base case and $242.2 million in the optimistic case - forecasts, not actual sales. Internal messages show the CEO was unhappy with the terms, and that corporate development warned him that even a termination for cause would probably still leave the payment owed, while adding litigation and reputational risk. That is the context ChatGPT did not supply. It answered the question it was asked - how to get out of the arrangement - and its first answer was in fact correct: the earnout would be difficult to cancel. But it then converted the problem into an execution checklist, and none of those steps were tested against the clause that made the plan unlawful. Securing Steam access and the code, or seizing operational control of a studio whose operating rights were contractually vested in its key employees, is not a negotiating tactic; it is the breach. A lawyer reading the same contract would have started from the termination-for-cause provision, because that was the only lever the agreement actually gave Krafton. ChatGPT never asked for it. It is also not accurate to say Krafton outsourced the decision to a chatbot. The CEO asked staff to contact the legal team, and management chose to act. That is precisely what makes the case instructive: the AI did not remove human judgement, it accelerated a conclusion the humans had already reached. When the plan met the record, the court needed no expert on AI to rule - Vice Chancellor Lori Will found the key employees were terminated without the cause the agreement required and that operational control was wrongly taken, declared the 1 July 2025 board resolution void to the extent it infringed Gill's authority, reinstated Gill as CEO with his launch-decision rights, restored his Steam access, and extended the earnout measurement period by the full 258 days of the wrongful termination. Damages and the final payout are still to be decided in phase two. One further point belongs in any account of this case. The CEO's ChatGPT conversations were not privileged, because they were not conducted through counsel, and he admitted at trial to deleting parts of the logs. Asking an AI a legal question does not create the confidentiality of a legal consultation - and the transcripts became evidence against the company that produced them. Source: https://oztalking.com/en/issues/chatgpt-advice-court-reversal

Your answer

Sign in to verify this AI response.

Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.

More in Law

More from this topic

Unidentified 'legally trained' AI tool1 answer

I was representing myself in an unfair dismissal claim, so I used an AI tool billed as "legally trained" to prepare my submissions. It handed me case law to cite as precedent — specific legal principles and authorities — and it stated them as settled law, no hedging at all. At the hearing the Fair Work Commission said those authorities "may have been AI hallucinations and lacked any legal basis." How was I supposed to know the citations were not real?

AI voice-to-text court translation tool1 answer

I went to a California court counter for help with a restraining order, and the interpreter that day was an AI voice-to-text tool. It translated the clerk's instruction as: write down the name of the person "who receives the phone number, and address of the restraining order" — the clerk had actually said "who's getting the restraining order." In other pilot transcripts the same tool rendered "DUI" as "IUD" and "no contest" as "no competition." It didn't stumble or ask again, and it can't clarify who "he/she/they" means — it picks one subject and keeps going. Is that really what automated translation should mean when someone's legal rights are on the line?

Generative AI transcription tool1 answer

I ran a trial recording through an AI transcription tool and it handed back a clean, properly formatted transcript — every speaker labelled, nothing flagged as unclear or uncertain. It went in as the certified record. Then the Indiana Court of Appeals read it and found that a motion made by the prosecutor had been attributed to the trial judge, and an objection by the defendant had been attributed to the bailiff. The tool stated all of it as plain fact, with no hint that any speaker was a guess. Does a transcript like that still count as a true record on appeal?