Does using ChatGPT or Claude waive attorney-client privilege?
A 2026 federal ruling is the first we know of to address the question directly. Here's what it held, what the ABA has said, and the practical questions for firms. This is general information, not legal advice.
Last updated October 2026
What the court held in United States v. Heppner
In a memorandum filed February 17, 2026, Judge Jed S. Rakoff of the Southern District of New York explained his ruling that documents recording a criminal defendant's exchanges with Claude, a public AI platform, were protected by neither attorney-client privilege nor the work product doctrine. The defendant had used the platform on his own, after receiving a grand jury subpoena, to prepare reports on his defense, and later shared them with his lawyers. The court called it a question of first impression.
The court's reasons
- The AI platform isn't a lawyer, so the exchanges weren't communications between client and attorney.
- They weren't confidential: the court pointed to the platform's consumer privacy policy, which said the provider collects inputs and outputs, may use them for training and may disclose data to third parties, including in litigation.
- The defendant wasn't seeking legal advice from the platform, and his lawyers hadn't directed him to use it. The court noted that if counsel had directed it, the analysis might be different.
- Sharing non-privileged material with lawyers afterward didn't make it privileged.
What it doesn't decide
The ruling concerns one defendant's own use of a public consumer AI service, in one federal district. It doesn't address a lawyer using AI at a client's direction, business plans with different terms, or AI running on the firm's own hardware. Talk to ethics counsel about your jurisdiction.
What the ABA has said
In ABA Formal Opinion 512 (July 2024), the ABA said lawyers using generative AI must protect client information under Model Rule 1.6, evaluate the risk of disclosure before entering client information, and in some cases get the client's informed consent. Running AI on hardware you own changes where data goes. It doesn't replace your own analysis under the rules of your jurisdiction.
Questions to answer before choosing a tool
- Where will client information be processed, and under whose terms?
- Can the provider keep, review or disclose what we enter?
- Can we keep matters separate, so the tool only answers from the right files?
- What should we tell clients about using AI tools themselves on their matters?
- Who checks AI drafts and citations before they're used?
Where private AI fits
Running AI on hardware your firm owns keeps prompts and documents inside your office rather than with a public AI service. That changes where data goes; it doesn't decide privilege questions, which stay with you. See private AI for law firms.