AI Use Impacts Attorney Client Privilege

AI Use Impacts Attorney Client Privilege

When I first read this recap of a recent court ruling, the outcome seemed obvious. And then I stopped and thought about it more carefully. There was an important shift in how we work with our clients that was easy to miss. I certainly had. 

William Galkin is a technology/cybersecurity/business attorney who consistently shares up-to-date legal information business leaders need to know. To avoid wandering into my own unqualified legal interpretation, I’ve included William’s post:

............................................................................................................................

"On February 17, 2026, the Southern District of New York issued a decision in United States v. Bradley Heppner, addressing a question many lawyers and executives have quietly debated:

If you use a public AI tool to analyze legal exposure, is that protected by attorney-client privilege?

The court said no.

In this securities fraud case, the defendant used a public generative AI chatbot to explore possible defenses and assess the government’s investigation. When authorities later seized records of those AI exchanges, he claimed they were shielded by attorney-client privilege and the work product doctrine.

Judge Rakoff rejected that argument.

Why Privilege Failed

• AI is not your lawyer.
Privilege requires a confidential relationship with a licensed attorney (or the attorney’s agent). A public AI system does not qualify.

• No reasonable expectation of privacy.
The platform’s published terms allowed collection and potential disclosure of user inputs and outputs. That undercut any claim of confidentiality — and disclosure to a third party can waive privilege.

• Counsel did not direct the AI use.
Because the defendant chose to use the tool independently, the AI was not treated as part of the legal team. The court focused on whether he was seeking advice from the AI — not from counsel.

The work product doctrine also did not apply because the materials were not created by or at the direction of an attorney.

What This Means for Companies

If executives or employees are entering sensitive legal analysis into public AI platforms, assume it could be discoverable.

Three practical takeaways:

1. Do not assume AI-assisted legal prep is privileged.
2. Review AI terms of use carefully.
3. Involve counsel before using AI in high-risk matters.

Public AI tools are powerful. But they are not protected legal advisors.
Privilege risk now belongs squarely inside your AI governance framework."

............................................................................................................................

The Implications About AI Use Are Unclear

The reasoning behind the judge’s ruling makes sense. We all know by now that sharing any information with a third party changes our expectation of privacy. We are the product, and our most valuable information is the currency we exchange for convenient access. 

My comment to William’s post included this question: If a client uses their attorney’s work product for AI research and analysis without their attorney’s explicit direction, is privilege lost. William’s response was “That’s the implication.”

The judge also said that had the attorney directed sharing the information with AI the outcome might have been different. That’s an important distinction.

The Conversation We Need to Have With Our Clients

AI disclosure policies are gradually becoming more commonplace for service providers in any field. Clients are asking how we use AI in our work for them, and we need to be transparent in our disclosure. That’s the new trust factor that influences our business relationships. 

What we need to include in our disclosure statement is not just how we use AI in our work. We need to be absolutely clear in how our clients can share with AI the work we create for them. 

Legal conversations are no longer the walled garden between attorney and client. Whether we know it or not, AI tools have taken a seat at the table. Clients want to feel in control of their outcome and asking their AI tool of choice to weigh in is empowering. Confidential information is unintentionally released into the wild.

Clients have relied on attorney-client privilege for so long that it’s only natural that they don’t consider the damaging impacts this new AI sharing can have. 

Every version of these AI tools has its own level of data control and protection. Instead of relying on these third parties to do the responsible thing (which we know isn't going to happen), service providers — not just attorneys — need to take an active role in protecting confidential information.

That means our disclosure wording should not only say what we do, but what our clients can do with the work we produce for them. Until I read William's post, I really hadn’t given any thought to this tool in the middle sand trap. Now it’s hard to unthink it. 

AI Use Isn't Just an Attorney-Client Problem

If we’re delivering non-legal work that comes from our knowledge and experience, we typically include copyright wording that’s intended to protect our intellectual property. No copying, reproducing, or sharing with a third party. But it might be time to revisit these last century words.

Third party disclosure often was assumed to mean shielding our work from a competitor. It was a human. Clients don’t think of ChatGPT, Claude and the SaaS products they use every day as third parties. They’re more like helpful tool friends, not an entity with a self-serving agenda.

Clarity Changes Old Thinking

The value of our data and the work we produce is only going to become more attractive to AI companies. Integration between AI and third party apps is becoming commonplace. As a result, our information footprint extends far beyond our awareness and control. 

For example, Claude Cowork promises to transform the drudgery of your daily life when you give it access to all your email, calendar, Slack conversation, and on and on. Copilot is right there doing the tedious work for all its Microsoft customers.

Access to your sacred data is a (terrifying) requirement. Sharing it with their undisclosed third party partners is baked into their Terms and Conditions. How much are you willing to sacrifice for superficial ease? 

The Bottom Line

My thought for you is this:

Privacy, confidentiality, and privilege need a new conversation that removes ambiguity and reframes long-standing expectations. We don’t want to wait for the court ruling to tell us our familiar rules have changed.