The CEO Used ChatGPT Instead of a Lawyer. The Court Used the Chat Log Against Him.
A Delaware Court of Chancery ruling just made one thing unmistakably clear: what you type into an AI tool can become the most damaging evidence in your case.
Most executives who use AI tools to think through a problem never imagine that those conversations will end up in a courtroom. They are not storing evidence. They are just thinking out loud — asking questions, exploring options, working through strategy the same way they might in a private notebook.
On March 16, 2026, the Delaware Court of Chancery made clear that this assumption is wrong. In Fortis Advisors, LLC v. Krafton, Inc., the court found that the CEO of a major gaming company had used ChatGPT to develop a strategy to avoid paying a $250 million earnout obligation following a corporate acquisition. He followed the chatbot's recommendations. The court followed the chat log — all the way to a finding of bad faith breach of contract.
The AI conversation the CEO thought was private became, in the court's words, a fully discoverable, searchable record of every prompt and every response — laying out the entire scheme from start to finish.
Fortis Advisors, LLC v. Krafton, Inc.
Krafton acquired Unknown Worlds Entertainment, the studio behind the game Subnautica, for a price that included a significant earnout — a contingent payment tied to post-closing performance targets. When it became clear the earnout would result in a substantial payout, Krafton's CEO turned to ChatGPT to develop a strategy to avoid paying it.
ChatGPT told him the earnout would be "difficult to cancel" — and then provided a strategy to do so anyway. He followed it. The company terminated the studio's key executives without contractually defined "Cause" and seized operational control of the studio. The court found all of it to be a deliberate, bad-faith breach.
Had he called a lawyer instead of a chatbot, that conversation would have been protected by attorney-client privilege. Instead, every prompt and every response was fully discoverable — and became the most damaging evidence in the case. The court reinstated the CEO, extended the earnout measurement period by 258 days, and issued broad injunctive relief.
AI Tools Are Third Parties. Courts Have Said So.
The Krafton decision is not an isolated ruling. It reflects a clear and consistent judicial position that has emerged across multiple jurisdictions in 2026: communications with AI platforms are communications with a third party. They are not protected by attorney-client privilege. They are not shielded by work product doctrine in most circumstances. And they are fully subject to the same preservation and production obligations as any other category of ESI.
For in-house legal departments, this creates a governance challenge that most organizations have not yet addressed. Executives, managers, and employees across every function are using AI tools to think through decisions — including decisions that may become relevant in litigation. The strategy session that used to happen with a lawyer now sometimes happens with a chatbot. The implications are not the same.
For law firms managing large matters, the question is operational: does your ESI identification process account for AI platform logs? Do your custodian interviews ask about AI tool usage? Do your legal hold notices instruct recipients to preserve their AI interaction history? If the answer to any of these is no, there is a gap — and the Krafton decision illustrates what that gap can cost.
Had he picked up the phone and called a lawyer, that conversation would have been protected by attorney-client privilege. Instead, his chatbot history produced a fully discoverable, searchable record of every prompt and every response.
Krafton Is One Case. The Pattern Is Much Larger.
Krafton sits alongside a growing body of 2026 decisions that collectively establish a framework for how AI interactions are treated in litigation. In Heppner, the Southern District of New York held that AI-generated materials created without attorney direction do not qualify for privilege. In Morgan v. V2X, the District of Colorado established that parties cannot upload confidential discovery data to consumer AI tools without contractual protections. And now Krafton demonstrates that AI chat logs can serve not just as ESI — but as the primary evidence of a party's intent and state of mind.
Courts are not treating AI tools as passive utilities. They are treating them as participants in decision-making — and the record those tools create is being used accordingly. Organizations that have not updated their information governance policies, legal hold frameworks, and custodian interview protocols to account for AI tool usage are operating with a gap that courts have already identified and are willing to exploit.
What Organizations Need to Understand Now
Five Things That Changed With Krafton
- AI chat logs are discoverable evidence of intent. Krafton established that AI conversations can be used to prove a party's state of mind — not just their data practices.
- There is no privilege for AI strategy sessions. Consulting an AI tool about a legal problem is not the same as consulting a lawyer. The protection does not transfer.
- Executives are creating ESI they don't know they're creating. Every AI prompt is a timestamped, searchable record. Organizations need to treat AI tool usage as an information governance issue.
- Legal hold notices must account for AI chat history. If litigation is anticipated, AI platform logs — where retained — are subject to the same preservation obligations as email.
- Custodian interviews need to ask about AI usage. Understanding which AI tools a custodian uses, for what purposes, and whether those tools retain history is now a baseline discovery question.
What Krafton reveals is not just a privilege gap — it is an information governance gap. Organizations have invested heavily in understanding where their email lives, how their documents are stored, and what their collaboration tools retain. Very few have asked the same questions about their AI tools.
The practical steps are not complicated, but they require intention. Audit which AI platforms your organization uses. Understand what each retains and for how long. Update your legal hold language to include AI interaction history. Add AI tool usage to your custodian interview framework. And ensure that employees who are using AI to think through decisions understand that those conversations may not be as private as they assume.
The Krafton CEO did not think he was creating evidence. He was just asking questions. That is the problem — and it is now your organization's problem to manage.
There is a simple rule emerging from 2026's AI discovery decisions: if you would not want opposing counsel to read it, do not type it into a chatbot. The privilege that protects a conversation with a lawyer does not follow you into the chat window.
Courts are not making this complicated. AI tools are third parties. Their logs are discovery. The organizations that understand this now will be the ones that avoid learning it the hard way.