In Fortis Advisors LLC v. Krafton, Inc. (Del. Ch. 2026), the Delaware Chancery Court ruled that a transcript of the gaming publisher's CEO using ChatGPT to avoid paying $250 million earnout to the developers of Subnautica 2 is evidence of bad faith. ChatGPT advised Krafton first that the earnout was noncancelable. After more prompts, it advised on a plan for Krafton to implement, which Krafton did. The court considered the AI transcripts as part of its bad faith analysis, and, in effect, the court ruled against Krafton in a way that Krafton tried to engineer with the chat bot to avoid. The case, in a way, is not about bad legal advice from AI. The issue is what happens when bad decisions based on AI output end up in court.
