The early wave of AI anxiety in the legal profession centred on hallucinations — fabricated citations, confident but incorrect statements of law, embarrassing courtroom moments. Those risks were visible and easy to understand. But a quieter and potentially more consequential risk category has been building in the background: what happens to attorney-client privilege and the work product doctrine when GenAI is woven into legal analysis and drafting?
That question has now landed in the courts.
In the tenth and final installment of his Regarding Evidence column in Today’s Managing Partner, iDS CEO & Founder Dan Regard examines two cases that are reshaping how legal professionals need to think about AI, confidentiality, and privilege: United States v. Heppner and Warner v. Gilbarco.
Two Cases, One Practical Warning
The cases arise from different contexts — one criminal, one civil — and they reached different outcomes. But Regard’s analysis makes clear that the more useful takeaway isn’t “one court got AI right and one got it wrong.” The outcomes are driven by doctrine and workflow facts, not by which AI model was involved.
In Heppner, a defendant used an AI tool for litigation preparation and the government sought to use those AI-assisted writings. The court rejected claims of both attorney-client privilege and work product protection, finding that the workflow didn’t satisfy the requirements for either — and that there was no reasonable expectation of confidentiality given how the account was configured and what the tool’s terms of service allowed.
In Warner, a pro se plaintiff used ChatGPT for litigation drafting and the defense tried to compel the prompts and related materials. The court denied that motion, affirming that work product doctrine isn’t limited to attorney work — it protects litigation preparation by or for a party — and that work product protections are not waived as easily as attorney-client privilege unless material has been disclosed to an adversary.
Four Problem Areas — and Four Answers
Regard distils the practical challenge into four problem areas that every legal team using GenAI needs to address:
Legal advice — AI is not counsel. Workflows must preserve the attorney’s role as the source of legal judgment, with counsel directing and reviewing work that involves theories, claims, and mental impressions.
Confidentiality and waiver — Disclosure to third-party AI systems can be fragile. The solution is enterprise-grade environments where training on client data is disabled, access is role-based, and retention is known and configurable. What matters isn’t whether a subscription is paid — it’s whether the environment keeps client strategy out of third-party reuse.
Content management — Most organizations cannot yet answer a basic question: where do your prompts live? AI outputs and prompts should be treated as potential ESI, with governance rules that address retention, deletion, and privileged repositories.
Hallucinations — No model reduces error to zero. Quality control must be built into the workflow: cite checking, quotation checking, and human review before any AI-assisted output becomes a filing.
The Bigger Picture
As Regard frames it, the issue isn’t AI in particular — it’s all modern writing environments and how an ever-evolving technology stack affects confidentiality, retention, and sharing. The path forward for outside counsel isn’t prohibition or reckless adoption. It’s informed, defensible use: doctrine-led, workflow-aware, and technically precise.
At iDS, navigating exactly that intersection is central to what we do. Our Information Governance, eDiscovery & Disclosure, and Cybersecurity practices help legal teams build the frameworks needed to use AI purposefully — without inadvertently surrendering the protections their clients depend on.
To connect with an iDS expert, visit idsinc.com.
iDS provides consultative data solutions to corporations and law firms around the world, giving them a decisive advantage – both in and out of the courtroom. iDS’s subject matter experts and data strategists specialize in finding solutions to complex evidence challenges, ensuring data can be leveraged as an asset, not a liability. To learn more, visit idsinc.com.
Having trouble with a technical term used in this post? Check out our Data Investigators Glossary to crack the code.
