Federal Courts Split on AI in Litigation: Privilege, Work Product, and Data Rules Diverge

By Hector Herrera | June 13, 2026 | Legal · Government Policy

Federal courts are issuing conflicting rulings on how generative AI tools interact with attorney-client privilege, work-product doctrine, and protective orders in active litigation — and the divergence is already creating a compliance nightmare for law firms and corporate legal departments. Some circuits now require AI platforms used in active cases to contractually guarantee zero data retention to preserve privilege. Others have set no standard at all. The patchwork is growing as AI adoption in litigation accelerates.

Attorney-client privilege protects confidential communications between lawyers and clients. Work-product doctrine protects lawyers' mental impressions, strategies, and case preparation from disclosure to opposing counsel. Both protections are foundational to how litigation works in the United States. When law firms upload case documents into AI tools to analyze evidence, draft briefs, or research precedent, those documents and the AI's outputs may enter data systems the law firm does not fully control — raising the question of whether the protections survive.

The Core Legal Problem

When you type sensitive case information into an AI tool, something happens on the other side: a company's servers process and, in many cases, store that data. The question courts are now being asked is whether that transmission — from attorney to AI platform — constitutes disclosure that waives privilege or work-product protection.

The stakes are high. A waiver finding means opposing counsel could potentially compel production of documents the law firm believed were protected. In a major commercial dispute or criminal case, that could be outcome-determinative.

According to Akin Gump's analysis of current circuit-level rulings, courts are splitting on three distinct questions:

1. Does using an AI tool waive privilege? Some courts have held that uploading privileged material into a commercial AI tool with data retention policies is equivalent to sharing it with a third party — which traditionally waives privilege under the "disclosure to a third party" exception. Other courts have applied a "common interest" or "functional equivalence" analysis, treating AI tools more like internal legal research software.

2. Does zero data retention preserve privilege? Several circuits are now conditioning privilege preservation on whether the AI vendor contractually guarantees that no case data is retained after the session ends. This is emerging as a de facto technical compliance requirement in some jurisdictions — use a retaining AI, lose your privilege argument.

3. How do AI tools interact with protective orders? In cases where courts issue protective orders limiting how parties can use confidential discovery material, some judges are now explicitly addressing whether uploading that material into AI tools violates the order's terms. Rules vary significantly.

What Law Firms and Corporate Legal Departments Face Now

The immediate practical problem is that the same AI tool — say, a contract analysis platform or a research assistant — may be safe to use in one federal district and potentially privilege-threatening in another. Litigators handling multi-district cases or national matters face a patchwork where no single policy works everywhere.

For corporate legal departments, the risk concentrates around in-house teams using commercial AI tools for document review, legal research, or contract analysis in connection with active disputes. If those tools retain data and the company is later sued or investigated, the question of what AI processed and stored becomes a discovery issue.

The practical compliance checklist that is emerging across firms and legal departments:

• Audit every AI tool currently used in litigation matters for data retention policies — get them in writing
• Classify matters by jurisdiction and apply circuit-specific standards when AI use is planned
• Contractually require zero data retention from AI vendors on any matter touching active litigation, even if your current circuit hasn't mandated it — because it may by the time the case concludes
• Brief partners and senior associates on the privilege-waiver risk in circuits that have already ruled

The Deeper Issue: Courts Are Reacting, Not Leading

What the Akin Gump analysis reveals is that courts are adjudicating AI privilege questions on a case-by-case basis without a consistent framework, because no such framework yet exists. The Federal Rules of Civil Procedure — the rules governing how federal litigation works — have not been updated to address AI. Local rules vary. Standing orders vary.

The American Bar Association has issued guidance encouraging but not requiring AI tool disclosure, and several state bars have issued ethics opinions, but none of these create binding rules in federal court. The result is that the common law is developing piecemeal, case by case, district by district — exactly what happened with email and electronic discovery in the late 1990s, except that AI adoption is moving much faster.

The eventual endpoint — uniform rules, either from the Judicial Conference or Congress — is not close. In the meantime, the patchwork is the compliance environment.

What to Watch

The Judicial Conference Advisory Committee on Evidence Rules has AI on its agenda. Watch for proposed amendments to Federal Rule of Evidence 502 (which governs privilege waiver) that could provide a national standard for AI tool use in litigation. Any such proposal would require years of comment and approval, but its emergence would signal that the federal judiciary is treating this as a systemic problem, not a series of isolated rulings.

Source: Akin Gump — Federal Courts Issue Diverging Rulings on the Use of Generative AI in the Context of Privilege, Work Product, and Protective Orders