There is a particular kind of shortcut that looks, from the outside, like diligence. You upload a stack of discovery responses into an AI tool, ask it to flag the deficient ones, and drop the results into a letter to opposing counsel and the court. The work got done quickly. The issues got identified. What could go wrong?
Quite a lot, it turns out. In White v. Walmart, Inc., No. 1:25-cv-01120 (S.D. Ind. Apr. 14, 2026), reported by eDiscovery Today, a federal magistrate judge sanctioned plaintiff's counsel for doing exactly that. The court found that counsel had uploaded the defendant's discovery responses into an AI program, asked the AI to identify insufficient responses, and then copied and pasted those results into a meet-and-confer email transmitted to both opposing counsel and the court. Magistrate Judge Tim A. Baker held that exclusive reliance on AI output — without any independent legal analysis — does not satisfy Rule 26's requirement to confer in good faith before bringing discovery disputes before the court. Sanctions followed.
The case is no longer a hypothetical. It is a data point. And for litigation teams already navigating a rapidly evolving body of AI-related case law, it should function as a compliance alarm.
What Rule 26 Actually Demands
Rule 26(g) requires that every discovery request, response, or objection be signed by an attorney who, after a reasonable inquiry, certifies that the submission is consistent with the rules, not interposed for an improper purpose, and neither unreasonable nor unduly burdensome. The meet-and-confer obligation under Rule 26(c) and the local rules that build on it demand something similarly substantive: that counsel actually engage with the dispute, evaluate the legal merits, and attempt a genuine resolution before seeking judicial intervention.
An AI tool, however sophisticated, does not perform that function. It pattern-matches. It identifies textual characteristics that resemble deficiencies. It cannot evaluate whether a given response is legally adequate under the applicable standard of production, whether a protective order might resolve the impasse, or whether the dispute is even worth pressing. Those judgments belong to the lawyer. When counsel substitutes the machine's output for that judgment — and transmits it to the court as if it were the product of legal analysis — the certification requirement is violated in substance, even if it is technically satisfied in form.
A Pattern Courts Are Beginning to Name
The April 2026 ruling does not stand alone. Morgan Lewis's Q1 2025 review of eDiscovery case law identified an accelerating judicial focus on two related themes: the demand for validation of AI-assisted workflows and the expectation of transparency when AI tools have been used in discovery. Courts across multiple jurisdictions were already scrutinizing whether AI-generated outputs had been reviewed and filtered by qualified counsel before being acted upon, and whether that review process was documented and available for inspection.
Read together, these developments reveal a coherent judicial posture. Courts are not hostile to AI-assisted discovery work. What they are hostile to is the use of AI as a liability shield — a way of claiming that work was done while avoiding the professional responsibility that the work entails. The attorney who uploads, prompts, and pastes has not done the lawyering. The attorney who reviews the output, applies legal judgment, corrects errors, makes strategic choices, and then signs a submission has done the lawyering. The distinction matters enormously, and courts are now drawing it explicitly.
Building a Defensible AI Workflow
The practical lesson from these cases is not that AI tools should be avoided. It is that their use must be structured so that counsel can demonstrate, if challenged, that no transmission to the court or opposing counsel was made on the basis of unreviewed AI output alone.
A defensible workflow requires several things. First, every AI-generated output that informs a court filing, discovery letter, or meet-and-confer communication should be reviewed by a licensed attorney who can attest to having independently evaluated its accuracy and legal adequacy. Second, that review should be documented contemporaneously — not reconstructed after a sanctions motion is filed. Third, firms should establish clear protocols specifying which tasks AI tools may assist with, what review standards apply to each category, and how discrepancies between AI output and attorney judgment are resolved and recorded. Fourth, where AI assistance has materially shaped a submission, counsel should consider whether transparency to the court is warranted either by local rule or as a matter of professional prudence.
The goal is not to generate paperwork. It is to ensure that the professional judgment required by Rule 26 actually occurs — and that it can be shown to have occurred.
How Law & Forensics Helps
Law & Forensics works with litigation teams to design and audit AI-assisted eDiscovery workflows that satisfy current judicial expectations for validation, transparency, and attorney oversight. From protocol development to expert analysis of disputed discovery processes, our forensic and legal technology specialists help counsel demonstrate that AI tools were used as instruments of professional judgment — not replacements for it. When sanctions risk is on the table, the difference between a documented review process and a copy-paste workflow can be dispositive.