Linked In Post: “One week. That’s the pace of AI and the law right now. Not months. Not years. Days”. Last week I wrote about United States v. Heppner……..
Chris SchwegmannManaging Partner at Lynn Pinker Hurst & Schwegmann
One week. That’s the pace of AI and the law right now. Not months. Not years. Days.
Last week I wrote about United States v. Heppner, where Judge Jed Rakoff held that a defendant’s Claude conversations were neither privileged nor work product — fully discoverable by federal prosecutors. I called it a five-alarm warning for corporate clients.
Seven days later, a federal magistrate judge in the Eastern District of Michigan reached a different conclusion on the work product question in Warner v. Gilbarco, Inc. — and the tension between these two opinions is something every litigator and GC needs to understand right now.
In Warner, Judge Anthony Patti refused to compel production of a civil plaintiff’s ChatGPT interactions, holding they were protected work product. His key move: treating AI as a tool, not a person. Under that analysis, using an AI platform doesn’t constitute disclosure to a third party in the way that matters for waiver — because work product waiver requires disclosure to an adversary, not merely to any third party.
That’s a meaningful doctrinal distinction from Heppner, where Rakoff treated Anthropic’s privacy policy as destroying any confidentiality expectation.
So where does that leave us? A few practical points:
? Attorney-client privilege remains essentially unavailable for AI communications. Both courts agree on that, and the reasoning is sound. Consumer AI platforms are not your lawyer.
? Work product protection is now genuinely contested. The two opinions reflect different theories of what the doctrine protects and whether AI use constitutes a waiver. Until appellate courts resolve this, you cannot counsel clients that their AI-assisted litigation analysis is safely protected.
? The facts still matter enormously. Heppner was a criminal defendant who acted entirely on his own initiative, without attorney direction. Warner was a civil plaintiff whose AI use looked more like traditional mental impression work product. Counsel-directed AI use — where the lawyer selects the tool, directs its use, and can demonstrate the analysis reflects legal strategy — presents the strongest case for protection under either framework.
? Enterprise tools with negotiated data protections remain meaningfully lower risk than consumer platforms. They address the confidentiality problem Rakoff focused on in Heppner, even if they don’t fully resolve the work product question.
The bottom line hasn’t changed: call your lawyer before you open the chatbot. But the legal landscape around what happens if you don’t is developing faster than anyone expected.
I’ve attached the Warner opinion for those who want to read it alongside Heppner. This is a space I’ll be watching closely.
One week. That’s the pace of AI and the law right now. Not months. Not years. Days.
Last week I wrote about United States v. Heppner, where Judge Jed Rakoff held that a defendant’s Claude conversations were neither privileged nor work product — fully discoverable by federal prosecutors. I called it a five-alarm warning for corporate clients.
Seven days later, a federal magistrate judge in the Eastern District of Michigan reached a different conclusion on the work product question in Warner v. Gilbarco, Inc. — and the tension between these two opinions is something every litigator and GC needs to understand right now.
In Warner, Judge Anthony Patti refused to compel production of a civil plaintiff’s ChatGPT interactions, holding they were protected work product. His key move: treating AI as a tool, not a person. Under that analysis, using an AI platform doesn’t constitute disclosure to a third party in the way that matters for waiver — because work product waiver requires disclosure to an adversary, not merely to any third party.
That’s a meaningful doctrinal distinction from Heppner, where Rakoff treated Anthropic’s privacy policy as destroying any confidentiality expectation.
So where does that leave us? A few practical points:
? Attorney-client privilege remains essentially unavailable for AI communications. Both courts agree on that, and the reasoning is sound. Consumer AI platforms are not your lawyer.
? Work product protection is now genuinely contested. The two opinions reflect different theories of what the doctrine protects and whether AI use constitutes a waiver. Until appellate courts resolve this, you cannot counsel clients that their AI-assisted litigation analysis is safely protected.
? The facts still matter enormously. Heppner was a criminal defendant who acted entirely on his own initiative, without attorney direction. Warner was a civil plaintiff whose AI use looked more like traditional mental impression work product. Counsel-directed AI use — where the lawyer selects the tool, directs its use, and can demonstrate the analysis reflects legal strategy — presents the strongest case for protection under either framework.
? Enterprise tools with negotiated data protections remain meaningfully lower risk than consumer platforms. They address the confidentiality problem Rakoff focused on in Heppner, even if they don’t fully resolve the work product question.
The bottom line hasn’t changed: call your lawyer before you open the chatbot. But the legal landscape around what happens if you don’t is developing faster than anyone expected.
I’ve attached the Warner opinion for those who want to read it alongside Heppner. This is a space I’ll be watching closely.