Using AI as your legal advisor can have devastating circumstances.

The recent decision in United States v. Heppner has sent a clear and sobering message through the legal community: using generative AI in legal matters without proper structure can lead to the destruction of privileged information. At issue was whether documents a defendant created through interactions with an AI platform were protected by attorney-client privilege or the work product doctrine. The court’s answer was unequivocal, 'no'. 

Those considering using AI for legal matters, especially those representing themselves (referred to as "pro se"), should know that long-standing legal principles still apply, regardless of the technology used. Here are three key takeaways every should understand.

1. No Privilege

The foundation of attorney-client privilege is a confidential communication between a client and a licensed attorney for the purpose of obtaining legal advice. In Heppner, that element was missing entirely.

The court made it clear: AI is not a lawyer. Conversations with an AI platform do not create a privileged relationship because there is no licensed professional bound by duties of confidentiality or loyalty.

Even if the information relates to legal strategy, it does not become privileged simply because it is later shared with an attorney. Privilege must exist at the time of the communication—not after the fact.

2. Voluntary Disclosure

Perhaps the most important and overlooked lesson is that entering information into a public AI platform is effectively a voluntary disclosure to a third party.

In Heppner, the defendant used a publicly available AI tool whose terms allowed for data collection and potential sharing. By doing so, he lost any reasonable expectation of confidentiality. And once confidentiality is lost, privilege is lost.

This principle is not new. What’s new is how easily people now share sensitive information without realizing it. Typing legal strategy, personal facts, or confidential business information into an AI tool is no different—legally—than sharing it with any other third party.

3. No “Agent” Status

Some argued that AI should be treated like an “agent” of the attorney similar to an investigator or consultant working under legal direction. The court rejected that argument.

In Heppner, the defendant used AI independently, without direction from counsel. As a result, the AI platform could not be considered an extension of the attorney-client relationship.

The implication is critical: for privilege to potentially apply, tools must be used under the supervision and direction of legal counsel, within a framework that preserves confidentiality. Without that structure, AI is just another third party.

The Bigger Picture

Heppner is not an anti-AI decision. It is a reminder that technology does not rewrite legal rules. Privilege depends on three things:

• A real attorney-client relationship
• Confidentiality
• Proper legal purpose

Remove any one of those, and the protection disappears.

As AI becomes embedded in everyday workflows, the risk is not just for lawyers, it’s for anyone who assumes that “asking AI” is private or protected.

It isn’t. And Heppner makes that clear.