AI Chatbot Transcripts Are Discoverable. Just Like Emails.

By now, we all know it’s a very bad idea for anyone in the legal profession to use AI as a shortcut to real legal work.  But did you know there could also be risks for defendants and plaintiffs using AI, too?

Your legal research is not automatically legally privileged.

On February 17, 2026, a federal judge ruled that a defendant’s AI-generated legal strategy research was 100% discoverable (United States v. Heppner).  It didn’t matter that he later shared that information with his attorneys.  The defendant created the research documents on his own and could not claim attorney/client privilege retroactively.

The judge also ruled that the AI-generated research was not covered by Work-Product Doctrine. Why? Even though the defendant created the documents intending to give them to his attorneys, he did not create them on advice of counsel.  And, he created them using a tool that had no expectation of privacy.

AI feels private – it’s NOT private.

For people who use consumer-grade AI to run scenarios and get perspectives on a variety of business applications, it just feels natural and easy to turn to AI to get legal analysis on a developing situation.

Don’t do it. 

Why? The judge in the Heppner case looked at the terms of service of the AI service the defendant used, and the judge saw that it clearly stated several things: 

• Their AI did not provide legal services and did not create an attorney/client relationship.
• All inputs and outputs would be logged.
• There was no guarantee of privacy.  In fact, the AI service stated that both input and output information would be used to train the AI model and might also be shared with third parties.

With no expectation of privacy, the judge determined there was then no expectation of confidentiality.

Consumer AI vs Enterprise AI

The defendant was using a consumer version of a popular AI service.  There are enterprise versions with different terms and conditions that have greater privacy protections.  Had the defendant used an enterprise version of an AI, some of the judge’s decisions based on expectation of privacy might have been different.

But enterprise AI was not in front of the judge, so that scenario is untested (as of this writing).  So, for now, assume that AI of any kind can NOT be treated the same as communication with or advice from legal counsel and should not be used as such.

Treat AI like discoverable emails and texts.

AI can bring vast amounts of information to you quickly – but AI is not your lawyer.  Think of AI as a third-party business consultant who can absolutely be subpoenaed to testify against you.

The United States v. Heppner judgment addresses AI documents created by the defendant.  But this judgment can affect plaintiffs, too.  The ruling could affect all kinds of litigation and legal work – civil suits, regulatory investigations, contract disputes, employee conflicts, even tax workflows.

What does AI use mean for your business and potential legal exposure?

The answer to this is going to depend on your situation.  Some things to ask:

• Do you provide AI tools to your employees?
• Do you let them use their own? (Would you know if they did?)
• Do you address AI use in your workplace policies? 
• Would anyone in your organization use AI to gather information in anticipation of regulatory or legal matters? (Would you know if they did?)

AI use adoption is growing steadily.  Companies need to think deeply about what that means and what kind of benefits – and exposure – AI use could bring.

_____________
 
Have questions about next steps?  We can help. 

If you think your business is not ready for the exposure AI can create, there’s no time like the present to fix that.

We would love to help you – call Scott A. Meyer at 469-916-7700 x115, or email him.