Can Your ChatGPT History Be Subpoenaed? A Federal Case Just Changed the Picture

Federal prosecutors introduced ChatGPT conversation logs as evidence in a criminal arson trial last week — and it wasn't a surprise legal maneuver. It was a logical extension of a trend that's been building since January: courts have decided that AI chatbot conversations are discoverable records, full stop.

If you use ChatGPT, Claude, or Gemini for work, this is worth understanding.

What happened in the Palisades trial

Jonathan Rinderknecht, 30, stood trial in federal court on charges of arson affecting interstate commerce related to a fire that federal prosecutors allege he set near Pacific Palisades on New Year's Day 2025 — a fire that merged into the catastrophic Palisades wildfire.

Among the evidence prosecutors presented: Rinderknecht's ChatGPT conversation history. The logs included AI-generated images of buildings engulfed in flames, messages expressing anger about wealth inequality and the ultra-rich, and — most directly — a screen recording showing Rinderknecht asking ChatGPT whether a person could be held legally responsible for a fire started by their cigarette. Prosecutors argued the logs showed intent and a fixation on fire and social grievance.

The jury deadlocked 10 to 2 in favor of acquittal. A mistrial was declared on June 26, 2026, with a retrial now scheduled for October 19.

The case matters beyond its outcome. Reporting from The Verge and Rolling Stone identified this as one of the first prominent federal criminal cases where AI chatbot logs were introduced as substantive evidence against a defendant. The legal machinery that made it possible had been tested months earlier.

The January precedent you may have missed

In January 2026, a federal judge in the Southern District of New York affirmed an order requiring OpenAI to produce 20 million de-identified ChatGPT user logs as part of copyright litigation brought by news organizations.

OpenAI had resisted — its Chief Strategy Officer Jason Kwon publicly argued the tech industry needed a new legal concept, "AI privilege," to protect user conversations from subpoena the way attorney-client communications are protected. The court rejected that argument. The logs, it ruled, were ordinary discoverable records. The fact that they were AI conversations, not emails or text messages, was irrelevant to their discoverability.

Bloomberg Law and the National Law Review both reported on the ruling's implications: until Congress creates a specific privilege for AI conversations — which it has not done — AI chat logs occupy the same legal space as enterprise software logs, emails, and any other stored electronic communication.

Why there is no "AI privilege"

Several existing legal privileges protect specific types of communication: attorney-client (legal advice), physician-patient (medical advice), spousal (marital communications), and a few others. These are narrow, carefully defined exceptions to the general rule that courts can compel disclosure of evidence.

AI chatbots don't fit any of them. ChatGPT is not your lawyer. Claude is not your doctor. Gemini is not your spouse.

OpenAI CEO Sam Altman has publicly acknowledged this gap, stating there is "no legal confidentiality" for ChatGPT conversations — that the company hasn't yet figured out how to provide users with meaningful confidentiality protections when talking to ChatGPT.

Under OpenAI's Government User Data Request Policy (effective January 1, 2026), the company will disclose basic account information in response to a valid subpoena, and will produce conversation content in response to a valid search warrant or court order. When legally permitted, OpenAI says it will notify affected users — but non-disclosure orders can prevent that notification.

What this means if you use AI for work

Most people using ChatGPT for work will never face a subpoena. But "most people" is not "everyone," and the relevant risk is about what you type, not just who you are.

Consider the range of scenarios where ChatGPT logs could become relevant in litigation or investigation:

• Employment disputes: If your company is sued and discovery covers AI tool usage, your work-account conversations may be in scope.
• Regulatory investigations: Agencies in financial services, healthcare, and legal industries can subpoena AI tool logs the same way they can subpoena emails.
• Civil litigation: In any case where your communications are relevant, opposing counsel can request them.
• Criminal investigations: As the Palisades case demonstrated, criminal prosecutors can and will use them.

The practical adjustment isn't to stop using AI tools — it's to treat them the way you treat email. You probably wouldn't paste a client's confidential health information into a personal Gmail. The same logic applies to AI.

A few specific steps:

Turn off chat history when you don't need it. ChatGPT's settings let you disable conversation storage, which prevents logs from being retained after a session ends. (OpenAI retains safety-related logs regardless, but disabling history significantly reduces what's stored long-term.)

Know what your business plan actually covers. ChatGPT Business and Enterprise accounts include business-data protections; Enterprise adds stronger retention, residency, and compliance controls. Contract terms do not block valid legal process, so review the agreement and retention settings that apply to your workspace rather than assuming a plan name makes records undiscoverable.

Don't paste specifics you wouldn't put in email. Client names, case numbers, financial details, unreported health information — these warrant the same care in a chatbot as in a message you're composing to a stranger.

Check your organization's AI policy. Many organizations have policies about which AI tools can be used with client data. Those policies exist partly for reasons like this.

The question everyone's avoiding

There's a version of this conversation that goes: "just use AI tools responsibly and you have nothing to worry about." That's largely true, but it dodges a harder question: most people don't know their AI conversations are stored, searchable, and potentially compelled from vendors by courts.

OpenAI, Anthropic, and Google all store conversation logs as part of how their services work. That's not a secret — it's in each company's privacy policy. But the legal implications of that storage have only recently become concrete through cases like the Palisades trial and the SDNY copyright order.

Now they're concrete. The professional response is to use AI tools deliberately — knowing that what you type is a record, and treating it accordingly.

Sources

• The Verge: "Prosecutors used ChatGPT logs as evidence in the Palisades fire trial"
• Rolling Stone: "Palisades Fire Suspect Allegedly Created Wildfire Images With ChatGPT"
• CNN: "The Palisades Fire arson trial ended in a jury deadlock and mistrial"
• ABC News: "Judge declares mistrial in Palisades Fire suspect's federal trial"
• Bloomberg Law: "OpenAI Must Turn Over 20 Million ChatGPT Logs, Judge Affirms"
• National Law Review: "When Chats Become Evidence: Court Affirms Order Requiring OpenAI to Produce 20 Million De-Identified ChatGPT Logs"
• OpenAI Government User Data Request Policy (effective January 1, 2026)