Bench & Bar September/October 2025

LAW PRACTICE MANAGEMENT

EVERYTHING YOU TELL CHATGPT COULD END UP BEING USED SAM ALTMAN'S WARNING:

AGAIN YOU

BY STEVE EMBRY

O penAI CEO Sam Altman recently said something that should grab every lawyer's attention, litigator or not. Appear ing on the Theo Von podcast, Altman warned: "So if you go talk to ChatGPT about your most sensitive stuff, and then there's … a lawsuit or whatever, … we could be required to produce that." THE DISCOVERY RISK That statement should send chills down the spine of practicing lawyers. It means anything and everything someone puts into ChatGPT or any other publicly facing large language model could be discoverable in litigation. Over the years, I've reviewed countless client documents, emails, and text messages only to find something that shouldn't be there. The story is always the same: "I didn't think anyone other than the recipient would ever see that." And that was for documents being exchanged with another person who at least had some relationship to the matter or sender. Then there were the file memos that people thought were private. I once had an engi neer tell me, when asked if he had produced all his documents, that yes, he had—except for materials in his "private" file at home. It was a good lesson in the mindset differences

between lawyers and clients when it comes to discoverability and privacy. THE GENAI PROBLEM But I shudder to think of all the material being fed into large language models today. People are encouraged to use these tools to brainstorm ideas, seek personal advice, draft documents, and share their secrets and they may be doing all that without consider ing that their inputs and outputs may be fair game for discovery down the road. That means their thought processes, what they considered and rejected, advice they didn't follow, offensive comments, emo tional outbursts, and strategic planning could all become discoverable. The impli cations are staggering. ATTORNEY WORK PRODUCT I also worry about how lawyers are using these tools. Many have encouraged attor neys to use GenAI to find weaknesses in opposing arguments, brainstorm case theories, and treat the technology like a first-year associate. But there's a critical difference between talking to an associate in your firm and communicating with a non-human entity over which you have no control. An entity

that may or may not keep your materials confidential and has no attorney-client rela tionship with your client. We've all been cautioned not to place client confidences into public LLMs. But what about attorney work product? The material that shows our thought processes, ideas we considered and rejected, theories of the case, witness assessments, and strate gic analysis? Have we waived work product protection? EVEN CHATGPT THINKS IT'S RISKY I asked ChatGPT itself about this issue. Its response was telling: "Placing thoughts and ideas into a large language model like ChatGPT could potentially waive work product protection. If a lawyer uses a public, consumer-facing LLM without a confi dentiality agreement or enterprise-level protections, inputting sensitive legal anal ysis might be considered disclosure to a third party." Imagine arguing to a court that your LLM use doesn't waive privilege, only to have opposing counsel stand up, smile smugly, and quote ChatGPT against you.

28 september/october 2025