Disclosure May Not Be Good Enough: What Judge Pittman’s AI Sanctions Warning Means for Litigators

A disclosure can satisfy a court rule. It cannot make an inaccurate citation, unsound legal analysis, or unsupported argument defensible.

That distinction is at the center of a warning from U.S. District Judge Mark Pittman in a noncitizen-voting case in the Northern District of Texas (United States v. Bhengura, No. 4:26-cr-00314). As Bloomberg Law and Business Insider reported, defense lawyers for Khalwinder Singh Bhengura — who was convicted of making false statements of citizenship to vote in a federal election — filed a motion seeking his release from detention pending sentencing. The motion disclosed that portions had been prepared with generative AI, as the local rule requires. Judge Pittman denied the motion and sharply criticized the filing, saying he intends to sanction the lawyers. A sanctions hearing had not yet taken place.

The dispute is a useful reminder for litigation teams: disclosure describes a process. It does not prove that the work product was checked, accurate, or sound.

Disclosure did not end the dispute

In a November 4 order denying the motion, Judge Pittman wrote that the Court was “disturbed by counsel’s repeated filings which subtly bend the law by proffering superficially persuasive but fundamentally meritless arguments so clearly generated by artificial intelligence.” He called the motion’s statutory reading “an abysmal reading of the statute” and said a citation was “so tone-deaf that only artificial intelligence could advance it.”

He also called the filing “lazy lawyering and a prime example of the inherent dangers of lawyers relying on artificial intelligence to draft their briefs,” and wrote that it raised “serious questions” about the three lawyers’ ability to practice law. According to the reporting, he indicated he would sanction the lawyers and warned that further conduct could result in a ban from practicing in the Northern District of Texas. Reporting on the scheduling of the sanctions hearing has varied.

The procedural status matters. As of the November 6 report, the lawyers had not yet been sanctioned; Judge Pittman had announced his intention to sanction them and scheduled a future hearing. The case is not a ruling that AI use itself is misconduct. It is a warning that disclosure does not shield lawyers from scrutiny of the work they submit.

Lead counsel Nicole Knox told Business Insider, “We have nothing to hide,” and said the team had followed the applicable rules, including the AI-disclosure requirement. Co-counsel Mario Herrera argued that the rule “does not prohibit the use of such tools, nor does it classify the use of such tools as misconduct.” He added, “for decades lawyers have used templates, forms, and other types of drafting software.”

That distinction is worth keeping clear. The question is not simply whether a tool was used. It is whether counsel can stand behind every argument, authority, quotation, and factual assertion in the filing.

Legal professional reviewing documents on a tablet

Courts are asking for verification, not just disclosure

Other recent developments point to the same practical concern: a statement that someone checked the work is not the same as a reliable process for checking it.

In Paulk v. Larva Labs, Judge Aileen Cannon sanctioned a pro se plaintiff whose filing certified that authorities had been “manually double-checked,” yet cited a nonexistent Supreme Court case. The judge ordered the plaintiff’s opposition to a motion to dismiss to be handwritten. The episode, reported by Reason, illustrates why certification language cannot substitute for examining the sources themselves.

Pennsylvania is considering a more formal requirement. A proposal published by the state Supreme Court’s AI committee would require all court filers, attorneys and self-represented parties alike, to certify that they reviewed AI-generated content for accuracy, including authorities, quotations, legal analysis, and facts. The proposal includes potential sanctions up to dismissal. It remains a proposal, not an adopted rule; comments are due November 24. The Pennsylvania Bulletin notice sets out the proposed language.

And in a different context, the Tenth Circuit upheld an attorney-fee award of about $115,000, rejecting the argument that the opposing lawyers’ use of AI alone made their billed hours unreasonable. The reported decision reinforces a narrower point: verification takes work, and a general claim that AI was used does not establish that particular hours were unreasonable.

These developments do not create one nationwide AI rule. Requirements vary by jurisdiction, and proposals are not binding rules. But the practical expectation is consistent: lawyers remain responsible for what they file. This is not an abstract risk. In Mississippi, two lawyers were barred from practicing in the district for two years after AI-related fabricated citations appeared in filings, as Reuters reported.

Keep a workflow you can explain

If a team uses AI to help prepare a filing, or to search, summarize, or organize exhibit material, it should be able to show what a qualified person reviewed and how. A defensible workflow is more than a final disclosure sentence. It should preserve the materials and steps needed to explain the work.

For legal filings, that means checking cited authorities in reliable sources, confirming quotations against the original text, testing summaries against the law, and verifying factual claims against the record. The responsible attorney, not the tool, must make the final judgment.

For exhibits, the same principle applies. A statement that a document was preserved is not as strong as a clear, consistent process that distinguishes the source from working copies and records how an exhibit was handled.

Preserve the exhibit and the record

The parallel matters in depositions and trial. Disclosure is an assertion, not a record. What protects the integrity of an exhibit is what the team can show: the original remained untouched, presentation annotations were made on a copy, and the version marked during a deposition was not changed after the deposition concluded.

If a different version is needed later, create a new copy. Preserve the deposition exhibit as marked. Keep the path from source to deposition to courtroom presentation clear.

The certified court reporter remains the keeper of the deposition record. AI can assist with search or organization, but it does not replace the reporter or the record. Testimony, the record, and exhibits must be preserved at all costs.

Attorneys and witness participating in a remote deposition

That discipline can also make deposition work more manageable. When teams can locate and present the intended exhibit without sorting through paper binders, duplicate files, or disconnected tools, there is less administrative friction, and less room for confusion about which version was used.

ExhibitSuite™ helps litigation teams prepare, present, annotate, share, and archive exhibits in one workflow. Teams can make presentation marks while preserving originals, manage when exhibits appear, and move case materials from deposition work toward trial. The goal is practical control, not a substitute for legal judgment or the certified reporter’s role.

For a closer look at how exhibit provenance connects to authenticity, see “When Anything Can Be Faked, Your Proof of Authenticity Becomes Evidence Too”.

Build evidence into the process

Judge Pittman’s warning is not that every use of AI will lead to sanctions. It is that disclosure alone may not answer the questions that matter when work is challenged: Was it checked? Can the reviewer explain how? Does the record support it?

For litigation teams, the same test applies to exhibits. A paragraph saying an original was preserved is less useful than a workflow that shows which file was the source, what annotations were added to a copy, and what version was marked during testimony.

Keep accountable professionals in control. Use AI as an assistant, not an authority. Preserve the record, testimony, and exhibits, and make the process clear enough to demonstrate, not merely describe.

If you’d like to see how ExhibitSuite™ can support exhibit handling from deposition through trial, request a demo.

Frequently asked questions about AI disclosure and sanctions

Did the lawyers in the Bhengura case violate the AI-disclosure rule?

No, according to the reporting. The motion disclosed that portions were prepared with help from a generative AI tool, as the local rule required. In United States v. Bhengura (N.D. Tex.), Judge Pittman denied the motion and criticized the quality of the filing — its legal reasoning and citation — not a failure to disclose.

Is disclosing AI use required in federal courts?

Requirements vary. Some courts and local rules now require disclosure when generative AI is used to prepare filings, while others rely on general professional-responsibility standards. Check the standing orders for the specific court before filing. Some proposed rules, such as Pennsylvania’s, remain proposals rather than adopted requirements.

What was the outcome in Paulk v. Larva Labs?

Judge Aileen Cannon sanctioned a pro se plaintiff whose filing certified that authorities had been “manually double-checked” but cited a nonexistent Supreme Court case. The court ordered the plaintiff’s opposition to a motion to dismiss to be handwritten.

Can lawyers use AI in legal work at all?

Nothing in the reported decisions prohibits lawyers from using AI. The consistent message is that lawyers remain responsible for verifying every authority, quotation, and factual assertion in their filings. Courts have treated verification as real work — the Tenth Circuit upheld a fee award that included it.

How does this apply to deposition exhibits?

The same principle: a statement that something was handled correctly is weaker than a process that can demonstrate it. For exhibits, that means originals preserved untouched, annotations made on copies, and the version marked during a deposition never altered after the deposition concluded.

Why do law firms use ExhibitSuite™ for this?

Because it builds the proof into the workflow. ExhibitSuite™ preserves the original document untouched while annotations and presentation marks are made on a separate version, keeps the exhibit record organized from deposition through trial, and sends exhibits to the court reporter in one flow. That means the answer to “can you show how this was handled?” is built into the platform, not reconstructed after the fact.