In a landmark decision handed down February 17, 2026, Senior United States District Judge Jed S. Rakoff of the Southern District of New York ruled that documents a criminal defendant generated using the publicly available AI platform Claude are not protected by attorney-client privilege and must be turned over to prosecutors. The ruling appears to be the first of its kind in the country and carries far-reaching consequences for anyone — in a criminal or civil case — who has used AI to help organize their legal strategy.
Key Takeaways
- Documents you create using publicly available AI platforms (such as ChatGPT, Claude, or Gemini) in connection with your legal case are likely discoverable by the opposing side.
- AI is not your lawyer. Communications with an AI platform do not receive the same legal protection as confidential communications with your attorney.
- Because AI platforms like Claude reserve the right to share user data with third parties and the government, courts have found users have no reasonable expectation of confidentiality.
- AI documents are not shielded simply because you later share them with your lawyer.
- However, if your lawyer directs you to use an AI tool as part of their legal strategy, the analysis may be different.
- Using AI to privately organize your own thoughts before emailing your attorney — if done carefully — may still be protected, but this ruling signals courts will scrutinize such use closely.
Background: The Case Before Judge Rakoff
The defendant in this case faces federal criminal charges including securities fraud and wire fraud. After his arrest, the government seized documents he had generated using Claude, the AI platform operated by Anthropic. His defense counsel argued those AI documents were privileged on three grounds: (1) the defendant had inputted information he learned from his lawyers into Claude; (2) he created the documents to assist in obtaining legal advice; and (3) he shared the resulting documents with his attorneys. Critically, however, his lawyers had never directed him to use Claude for this purpose.
What Is Attorney-Client Privilege — and Why Didn’t It Apply Here?
Attorney-client privilege is a foundational protection in our legal system. It shields from disclosure (1) communications between a client and their attorney, (2) that are intended to be and kept confidential, and (3) made for the purpose of obtaining legal advice. The privilege is intentionally narrow: courts require broad disclosure of relevant evidence, which is why items like journals, diaries, and social media posts can be produced to the opposing side.
Judge Rakoff found that the AI documents failed to satisfy at least two, and likely all three, of these elements.
1. AI Is Not Your Attorney
The first element requires a communication between a client and their lawyer. Claude is not an attorney. The court found that communications between two non-attorneys simply cannot be privileged, no matter how legal the subject matter is.
2. No Reasonable Expectation of Confidentiality
The second element requires that the communication be kept confidential. The court pointed directly to Claude’s privacy policy, which notifies users that Anthropic may disclose data to third parties, including the government, in connection with legal claims, disputes, or litigation, and to train its AI models. Because that disclosure risk is baked into the terms of service, the defendant had no reasonable expectation of privacy in his Claude sessions. In effect, when you type sensitive legal information into a publicly available AI platform, you are sharing that information with a third party, which courts have consistently held destroys confidentiality and strips away any privilege.
3. Seeking Advice From AI Is Not the Same as Seeking Advice From Your Lawyer
The third element — which the court acknowledged was a closer call — requires that the communication be made for the purpose of obtaining legal advice from counsel. Here, the defendant was seeking legal guidance from Claude itself, not from his attorneys. The fact that he later forwarded the AI-generated documents to his lawyers did not retroactively privilege them. The court was clear: the act of relaying AI output to your attorney does not transform that output into a protected attorney-client communication.
The Work Product Doctrine Also Provides No Protection
The defendant also argued that the AI documents were protected under the work product doctrine, which shields documents created by or at the direction of an attorney in anticipation of litigation. Judge Rakoff rejected this argument as well. Because the defendant created the Claude documents entirely on his own initiative and not at the behest of his counsel, they do not qualify as attorney work product. This is a meaningful distinction: the work product doctrine is designed to protect the attorney’s thought process and litigation strategy, not a client’s independent use of a consumer AI tool.
The Important Exception: When Your Lawyer Directs You to Use AI
Judge Rakoff’s opinion left open an important distinction that clients and practitioners should note. The court acknowledged that if the defendant had used Claude at the direction of his counsel, the AI could have functioned as something akin to a highly trained professional acting as the lawyer’s agent which is within the established scope of attorney-client privilege. In other words, AI used as a tool by your attorney, or under your attorney’s specific instruction as part of their legal strategy, may be different than AI you use independently to brainstorm your own defense.
Additionally, the court left open the question of whether AI tools deployed within a “closed universe”. For example, a secure, confidential platform specifically designed for attorney-client work rather than a publicly available consumer product might warrant different treatment. This is an evolving area of law.
What Does This Mean for Civil Litigation?
While this ruling arose in a criminal case, the attorney-client privilege and work product doctrine apply equally in civil litigation including employment discrimination cases, wrongful termination claims, and harassment matters. If you have used publicly available AI to strategize about your case, draft your own account of events, or formulate arguments, that material may be discoverable by your employer’s lawyers. The legal principles the court applied are not limited to criminal proceedings.
Frequently Asked Questions
Are my AI chat sessions protected by attorney-client privilege?
Based on this new ruling, no. A federal court has held that communications with publicly available AI platforms like Claude, ChatGPT, or similar tools are not protected by attorney-client privilege. Because these platforms reserve the right to share user data with third parties, users have no reasonable expectation of confidentiality which is a prerequisite for privilege.
Can the other side obtain my AI-generated documents in a lawsuit?
Yes, according to this ruling. Documents you generated on a publicly available AI platform in connection with your legal matter are discoverable by opposing counsel. You should discuss any AI use with your attorney immediately.
What if I share my AI documents with my attorney — are they then protected?
No. Judge Rakoff specifically ruled that forwarding AI-generated documents to your lawyer does not retroactively privilege them. The privilege analysis focuses on the nature of the original communication, not what you do with it afterward.
Is it ever safe to use AI in connection with my legal case?
Potentially, if your attorney specifically directs you to use an AI tool as part of their litigation strategy. The court suggested that lawyer-directed AI use may be treated differently, more like using a professional agent employed by your attorney. Questions also remain about confidential, closed-system AI tools not available to the general public. When in doubt, ask your attorney before using any AI in connection with your case.
The Bottom Line
Judge Rakoff’s ruling is a watershed moment in the intersection of artificial intelligence and the law. It sends a clear message: publicly available AI platforms are not confidential, and using them to strategize about your legal case is not the same as speaking privately with your attorney. Clients should avoid using consumer AI tools including ChatGPT, Claude, Gemini, or any similar platform to work through their legal strategy without first consulting their lawyer. The opposing side may now have a road map directly into your thought process.
If you have an employment law matter and are concerned about how you’ve been using AI tools, or if you have questions about privilege and confidentiality, contact Phillips & Associates today for a confidential consultation. We represent employees exclusively, and our only interest is protecting yours.