AI-generated legal work exists in a gray zone. The Heppner and Gilbarco cases reached conflicting conclusions on whether AI-assisted work product qualifies for protection — meaning your firm's AI use policy could determine the outcome of your next privilege dispute.
What Happened at Legalweek 2026
Two federal courts looked at the same question — does attorney-client privilege and the work product doctrine protect documents created with AI assistance — and came to opposite conclusions.
In Heppner, the court found that AI-generated analysis could qualify for work product protection when an attorney meaningfully directed the AI's output and applied independent legal judgment to the result. The AI was treated as a sophisticated research tool — no different in principle than a Westlaw search.
In Gilbarco, the court took a harder line. It found that because the AI model processed client information and generated substantive legal conclusions without direct attorney authorship, the resulting documents were not entitled to work product protection. The court's concern: the "mental impressions" required for work product protection belong to an attorney, not a machine.
The split means there is currently no clear federal rule. And it means your firm's AI workflow — specifically how attorneys interact with AI outputs — may now be litigation-relevant.
Why This Matters Right Now
Think of work product privilege like a shield on a sword. The sword is your legal strategy. The shield protects your preparation materials from discovery. The Gilbarco ruling says: if a machine made the shield, it might not protect the sword.
For law firms deploying legal AI at scale, this creates three immediate exposure points:
1. Discovery requests targeting AI-generated documents. Opposing counsel now has a roadmap to argue that AI-assisted work product isn't protected. Expect this argument in discovery disputes throughout 2026.
2. Malpractice exposure from unreviewed AI output. If an AI-generated document isn't protected and it contains errors an attorney didn't catch, you've potentially got both a discovery problem and a malpractice problem.
3. Client disclosure obligations. Some state bar ethics opinions now require disclosure when AI tools are used in client matters. The Gilbarco ruling adds urgency to those policies.
What the Heppner Standard Requires
The Heppner court's more favorable interpretation — that AI-assisted work can be protected — came with a clear condition: the attorney must be doing the thinking, not just accepting the output.
Specifically, the court looked for:
- The attorney directed what the AI analyzed and why
- The attorney reviewed, revised, and applied independent judgment to the output
- The attorney's mental impressions are reflected in how the final work product was shaped — not just what the AI generated
This is actually a reasonable standard — and it's achievable with the right workflow. But it requires intentional practice, not default usage.
What Your Firm Should Do This Week
Step 1: Audit your AI use cases. Which matters involve AI-generated documents that could be subject to discovery? Deposition prep, research memos, timeline analysis, document review summaries — all potential targets.
Step 2: Establish an AI documentation protocol. For every AI-assisted document, attorneys should maintain a short record of: what prompt was used, what the AI produced, and how the attorney modified or applied judgment to the output. This record supports the Heppner standard.
Step 3: Update your engagement letters. If you haven't already, your standard engagement letter should address AI tool usage. Some firms are adding AI disclosure language proactively; others are waiting for state bar guidance. The Gilbarco ruling suggests waiting carries risk.
Step 4: Choose AI tools that support your privilege defense. Legal AI platforms purpose-built for attorney use — with prompt history, attorney review workflows, and audit trails — are now risk management tools, not just productivity tools.
The Bottom Line
The work product doctrine was written for a world where attorneys did all the thinking. AI changes that. The courts are now drawing the line at the same place any sensible risk manager would: did a lawyer actually think about this, or did they just hit accept?
For firms using legal AI well — with clear human oversight, documented review, and genuine attorney judgment — the Heppner standard is achievable. For firms treating legal AI as an autocomplete button, Gilbarco is a warning shot.
VerdictLegal is designed for the Heppner standard. Every research output, citation, and document draft is structured for attorney review — not attorney replacement.
VerdictLegal provides AI-assisted legal research, document drafting, and citation verification. This post is for informational purposes and does not constitute legal advice.
