Privileged Until You Hit Send: What a Landmark AI Ruling Means for Your Business
By Agustin Bauer
If you have ever typed confidential business information, deal terms, or legal strategy into ChatGPT, Claude, or a similar AI tool, a federal court just signaled that those conversations may not be protected. Even if you later shared them with your lawyer. For anyone who uses AI in connection with legal matters, transactions, or sensitive business decisions, this ruling deserves your attention.
What Happened
In United States v. Heppner (S.D.N.Y., February 17, 2026), Judge Jed S. Rakoff addressed what he described as a question of first impression nationwide. Bradley Heppner, facing federal fraud charges, used Claude (a publicly available AI platform) on his own initiative to develop defense strategies and analyze the legal issues in his case. He did this without his lawyers directing him to do so. He later shared the AI-generated documents with his legal counsel.
When the government seized those documents during a search, Heppner’s lawyers argued they were protected by attorney-client privilege and the work product doctrine. The court disagreed on every ground.
Why the Court Said No
The court found the privilege failed on two independently dispositive grounds.
First, AI is not an attorney. Attorney-client privilege protects communications between a client and a licensed lawyer who owes fiduciary duties and is subject to professional discipline. No such relationship exists, or can exist, between a user and an AI platform.
Second, the communications were not confidential. The court examined the AI platform’s privacy policy and found that it permits the collection of both user inputs and outputs, allows use of that data for model training, and permits disclosure to governmental authorities and in connection with legal proceedings. Under those terms, the court found no reasonable expectation of confidentiality in Heppner’s communications with the platform.
The court also found that Heppner was not seeking legal advice from the AI (the platform itself disclaims providing legal advice) and was not directed by counsel to use it. Although Heppner’s lawyers argued he used the AI for the “express purpose of talking to counsel,” the court drew a clear line: what matters is whether the user intended to obtain legal advice from the AI, not whether he later forwarded its output to his attorney.
On the work product side, the court reached the same conclusion. Even assuming the documents were prepared in anticipation of litigation, they were not prepared by or at the direction of counsel, and they did not reflect counsel’s legal strategy at the time they were created.
Critically, the court emphasized that information which starts out privileged loses that protection once shared with a consumer AI platform. Sharing previously privileged material with a third party, which is what an AI platform is, constitutes a waiver that cannot be undone by later routing the material through your lawyer.
A Necessary Word of Context
This is a single district court ruling, not binding nationwide precedent. The court itself acknowledged that different facts could produce a different outcome, for example if counsel had directed the AI use, or if a secure enterprise platform with contractual confidentiality protections had been used. Legal commentators have noted that the opinion’s reasoning is tightly tied to its facts, and future courts may refine, expand, or distinguish it.
That said, the underlying legal principles are well established and apply broadly across jurisdictions. This ruling is widely expected to be the first in a series of decisions shaping how courts treat AI use in legal contexts.
Beyond the Courtroom: Trends to Watch
This ruling matters beyond privilege. It should prompt your business to reexamine how AI tools interact with sensitive information across several areas that we see regularly in our practice.
AI note-takers on calls with counsel. Tools like Otter.ai, Fireflies, and Zoom’s AI assistant are increasingly recording and transcribing meetings by default, including calls with your attorneys. Those transcripts are stored on third-party servers, and the providers’ terms often permit data collection, retention, and use for model improvement. Under the reasoning in Heppner, allowing an AI note-taker to record a privileged conversation could compromise the confidentiality that privilege requires. Clients should disable these tools before any call involving legal advice.
Feeding deal terms and draft agreements into consumer AI. We regularly see clients paste term sheets, letters of intent, draft contracts, or counterparty proposals into AI chatbots to get a quick initial analysis. That exposes confidential deal terms, negotiation positions, and potentially proprietary information to a third-party platform. Under this ruling’s framework, that information could become discoverable in litigation, and any trade secret protection you assumed you had may be weakened if you cannot demonstrate reasonable measures to maintain secrecy.
The consumer vs. enterprise distinction. Not all AI platforms carry the same risk, and the distinction matters practically. Enterprise-grade AI platforms offered under commercial agreements often include contractual confidentiality protections, prohibit training on user data by default, and in some cases offer Zero Data Retention (ZDR) arrangements where inputs and outputs are not stored beyond immediate processing needs. Consumer-tier platforms (including free and individual paid plans) allow users to opt out of model training, but opting out does not necessarily prevent the platform from retaining or disclosing data under its privacy terms. It reduces one dimension of risk, but it is not a safe harbor.
Enterprise tools strengthen the argument that confidentiality was maintained, which is one of the elements the Heppner court found lacking. However, enterprise tools do not create privilege on their own. The AI still is not your lawyer, and using it without counsel’s direction still falls outside the protected relationship. The safest position remains using AI tools that have been vetted by your attorney, under your attorney’s direction, as part of a structured legal workflow.
What your Business Should Do Now
Talk to your attorney before using any AI tool in connection with a legal matter, transaction, or dispute. The Heppner court suggested that attorney-directed AI use might receive different treatment. Make that conversation happen before you open the chatbot, not after.
Disable AI recording and note-taking features on calls involving legal counsel. Check your default settings on Zoom, Teams, and any third-party transcription tools. If an AI assistant is joining your calls with your lawyer, you may be creating discoverable records of privileged conversations.
Do not input confidential deal terms, privileged communications, or legal strategy into consumer AI platforms. This includes term sheets, draft agreements, counterparty proposals, and any information received from your attorney.
Ask your counsel about enterprise-grade AI platforms with appropriate confidentiality protections. If your business needs to use AI tools operationally, the right platform selection, combined with attorney oversight, can meaningfully reduce the risk of inadvertent privilege waiver.
This article is for informational purposes only and does not constitute legal advice. If you have questions about how AI use may affect your legal protections, we encourage you to reach out to our team.