AI may be the most powerful strategic thinking tool executives have ever had. It may also become one of the most revealing sources of evidence in litigation.
CEOs and CFOs now use ChatGPT, Claude and other AI tools to examine contracts, disputes, M&A terms, financial exposure and difficult personnel decisions. These conversations can contain the same sensitive material that executives would traditionally discuss with legal counsel. Yet a private AI conversation is not automatically a privileged legal communication.
In 2026, courts began confronting that distinction directly. Their decisions show why the issue is moving quickly, and why businesses cannot wait for a single universal rule.
Four decisions, different answers
In United States v. Heppner (Southern District of New York, February 17), a court rejected attorney-client privilege and work-product protection for a criminal defendant's conversations with a consumer version of Claude. The defendant consulted AI independently and later shared some output with his lawyers. The court emphasized the absence of attorney direction and the confidentiality implications of the platform's terms.
In Warner v. Gilbarco (Eastern District of Michigan, February 10) and Morgan v. V2X (District of Colorado, March 30), courts instead recognized work-product protection for certain self-represented litigants' AI-assisted litigation preparation. Those outcomes were fact-specific, not blanket protection for all AI chats.
Most recently, on October 21, Fontainebleau Florida Hotel v. Botach addressed AI chat discovery and proposed subpoenas to four AI service providers. The Southern District of Florida held that AI prompts and outputs can qualify as work product even without lawyer direction, but rejected blanket objections. Protected material must be justified document by document, while the parties must address how to protect claims when records are sought from providers. This was a discovery ruling, not a universal precedent.
The distinction matters: attorney-client privilege and work-product protection are separate doctrines. Neither arises automatically from mentioning legal issues to a chatbot.
When AI becomes evidence of executive intent
The corporate exposure extends beyond legal research. In March, the Delaware Court of Chancery considered a CEO's ChatGPT exchanges in Fortis Advisors v. Krafton, a dispute involving a potential $250 million acquisition earnout. The court examined an AI-assisted takeover strategy among the evidence and found contractual breaches. The case did not create a general privilege rule, but it demonstrated how AI records can illuminate a decision-maker's reasoning.
Think about the practical questions: What was management trying to accomplish? What alternatives were considered? What risks were acknowledged? What instructions did an executive provide before taking action?
For finance leaders, those questions touch restructuring, lender relationships, financial reporting, acquisitions, earnouts, investigations, and employment decisions. AI may preserve a remarkably detailed draft record of how a decision evolved.
Privacy is not privilege. Deletion is not a strategy.
Enterprise confidentiality settings, restricted data use and appropriate retention controls can strengthen a company's position, but none alone creates attorney-client privilege. A conversation with a model is not the same as a confidential communication with counsel. Attorney-directed work and qualifying litigation preparation require careful factual analysis.
Nor should organizations respond by indiscriminately deleting history. When litigation is pending or reasonably anticipated, relevant information may have to be preserved. Deletion after a preservation duty arises can create a separate dispute about spoliation.
What leadership should change now
- Define sensitive use cases. Identify legal strategy, investigations, employment disputes, transaction negotiations and regulated financial matters that require special treatment.
- Involve counsel early. Where litigation or legal advice is at issue, let counsel determine the appropriate AI workflow and documentation. Do not assume after-the-fact sharing cures a privilege issue.
- Know the provider terms. Evaluate retention, data use, confidentiality, administrators, subprocessors and third-party access before placing sensitive information in an AI service.
- Plan for discovery and holds. Know which accounts, agents, transcripts and connected applications may retain relevant records. Preserve when required, with documented access controls.
- Separate analysis from authority. AI can assist the thinking; management and counsel remain responsible for advice, decisions and professional judgment.
AI is becoming part of how companies think. Courts are now deciding how much of that thinking may become evidence.
Executives should understand those boundaries before their next sensitive conversation, not after the opposing party requests the transcript.
Sources and legal context
- Fontainebleau Florida Hotel v. Botach, October 21, 2026 discovery order
- Debevoise & Plimpton, October 28 analysis of Fontainebleau and Heppner
- Harvard Law School Forum / Fried Frank, 2026 AI privilege cases
- Fortis Advisors v. Krafton, Delaware Court of Chancery, March 16, 2026
- Washington State Bar Association, AI-enabled tools advisory opinion 2025-05
Executive and technology analysis only, not legal advice. Privilege, work product, retention and discovery obligations are jurisdiction- and fact-specific. Consult legal counsel.