Practice Areas
“I Know AI When I See It”: But What Will the Courts See?

In 1964, Supreme Court Justice Potter Stewart famously wrote, “I know it when I see it,” in Jacobellis v. Ohio, 378 U.S. 184 (1964). He was talking about obscenity. More than sixty years later, I find myself recalling that observation for an entirely different reason: artificial intelligence.
These days, I sometimes feel that I know AI when I see it. Emails that once consisted of three sentences now arrive as three-page legal memoranda. A simple contract dispute suddenly produces lengthy discussions of constitutional rights, statutory violations, and legal remedies. People who have never practiced law are citing obscure cases, quoting statutes, and using legal terminology that often has little to do with the dispute. Sometimes the cases do not even exist.
Then there are the repetitive arguments, elaborate explanations, and impressive-sounding conclusions that say remarkably little. Of course, lawyers have been guilty of that for generations, so perhaps we should not be too quick to blame AI.
The real danger may not be what AI tells you. It may be what you tell AI.
Consider a client involved in litigation or under government investigation who uploads confidential emails, business records, communications with counsel, or other sensitive documents into an AI platform. Perhaps the client asks how to explain a questionable transaction, respond to an FBI inquiry, or defend against damaging allegations. That conversation may feel private. But is it privileged? And can an opposing attorney or government investigator obtain it?
Two New York decisions illustrate why these questions matter.
In United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), a defendant facing a federal criminal investigation used the Claude AI platform to explore possible defenses. The FBI seized approximately 31 AI-generated documents. Although the defendant had retained counsel and later shared the materials with his attorneys, Judge Jed Rakoff ruled that the documents were not protected by attorney-client privilege or the work-product doctrine. The defendant had independently created the materials using a third-party AI platform. Sharing them with his lawyers afterward did not make them privileged.
A few months later, however, a New York state court reached a different result. In Assini v. Hayward, 2026 NY Slip Op 26086 (Sup. Ct. Nassau County 2026), the court quashed subpoenas seeking a self-represented litigant's AI prompts, uploaded materials, and generated responses. The court recognized that AI-assisted litigation preparation may qualify for work-product protection.
The decisions involved different circumstances, and neither establishes a blanket rule. But they demonstrate that courts are already confronting whether AI conversations and the materials behind them can be discovered.
At our firm, we increasingly include requests for relevant, nonprivileged AI prompts, uploads, and generated materials in discovery where appropriate. After all, if someone used AI to prepare a factual statement, analyze documents, or develop a position in litigation, what they originally told the system may be just as important as the final document they produced.
That does not mean every AI conversation is discoverable. Privilege, work-product protection, relevance, and other discovery limitations still apply. But these are now real issues that clients and lawyers need to consider.
If you are involved in litigation, facing a government investigation, or dealing with a sensitive legal matter, speak with your attorney before uploading case-related documents or discussing the facts with an AI platform. And if you have already done so, do not delete those records without first discussing your preservation obligations with counsel.
Lawyers should also be warning their clients about these risks at the beginning of a representation. AI is here to stay, and I expect it will become an increasingly useful part of legal practice. I certainly do not suggest that people stop using it. But there is a difference between asking AI to explain a legal concept and asking it to analyze confidential documents or help you explain potentially damaging evidence.
Justice Stewart said he knew it when he saw it. Today, the more important question may be what a judge, prosecutor, or opposing counsel will see when they obtain your AI conversations.
And that is something worth thinking about before you hit “Enter.”
THE BOIES LAW FIRM is located in the heart of New York City. We focus on complex litigation, investigation, corporate practice, and civil rights. Contact us at: (212) 729-0085 or contact@boieslaw.com
