When Courts Punish AI Hallucinations

If your firm runs AI tools on legal research and drafting, you are ahead of most litigators on the North Shore. But if you stop before verification, you are carrying a risk that federal courts are now pricing in fines, suspensions, and disciplinary referrals.

The cases are not hypothetical. Between 2023 and mid-2026, more than a dozen attorneys have faced sanctions for submitting AI-generated briefs containing citations that never existed or wildly misrepresented the holdings they were supposed to support. The courts are not sympathetic, and the consequences are getting more severe.

The Case Docket Builds Fast

The signal case is Mata v. Avianca, Inc. (S.D.N.Y., 2023). Roberto Mata filed a personal injury suit against the airline. His attorneys, without verifying the output, filed a brief built on ChatGPT research. The brief cited cases that did not exist, quoted holdings that were invented, and named judges who had no record of writing the opinions cited. Judge P. Kevin Castel fined the attorneys and their firm $5,000 and required them to write letters of apology to the judges whose names ChatGPT had fabricated. The legal press called it a cautionary tale. Courts called it a starting point.

Since Mata, the sanctions have grown larger and the consequences more serious. According to the ABA Journal, the 6th U.S. Circuit Court of Appeals issued a combined $30,000 sanction against two attorneys in 2025 for submitting briefs containing more than two dozen fabricated citations. The court dismissed the case entirely, finding the pervasive misconduct rendered it almost entirely frivolous.

In June 2026, the 9th U.S. Circuit Court of Appeals sanctioned two California immigration attorneys, Mike Sethi and William Rounds, $2,500 each and suspended both from practicing before the 9th Circuit for six months after they filed briefs containing AI-generated errors and then initially blamed "typographical mistakes" while denying AI involvement. The court also required them to disclose AI use in all filings for two years. The court found that at every subsequent step, the attorneys knowingly or recklessly made false statements. Hiding the source of the error cost them far more than the error itself.

A Northern District of Georgia court in Boston v. Williams (2025) found that 17 of 24 citations in the plaintiff's summary judgment opposition were either inaccurate or nonexistent, sanctioning the attorney under Rule 11. A Southern District of Ohio court imposed $7,500 against two attorneys plus a contempt finding and a referral to the state disciplinary board for Rule 11 violations the judge described as the most egregious he had seen on this subject.

What Rule 11 Actually Requires of You

Federal Rule of Civil Procedure 11(b) requires that every attorney who signs and files a document with the court certifies that the legal contentions in it are warranted by existing law or by a nonfrivolous argument for changing the law. That obligation sits on the attorney, not the AI tool.

Every court that has addressed this has said the same thing: there is nothing in Rule 11 that prohibits using AI to draft briefs. What Rule 11 prohibits is signing and filing a document you have not verified. If you ask Claude or ChatGPT to draft a motion and file it without confirming each cited case exists, says what you attributed to it, and supports the proposition you are advancing, you have signed a document you did not verify. That is the violation.

The courts have identified a consistent aggravating factor: candor. In Mata, the attorneys initially maintained the brief's accuracy even when opposing counsel could not locate the cases cited. In the 9th Circuit case, the attorneys' initial denial of AI involvement, followed by a later admission it was "probably" involved, drove the court toward suspension rather than a lesser remedy. The ABA Journal's reporting on candor and AI sanctions shows a clear pattern: prompt disclosure and correction reduce penalties. Concealment multiplies them.

For a Lake Forest or Evanston litigator practicing in federal court, the risk profile is the same as in any other jurisdiction. The 7th Circuit had not issued a major AI sanctions order as of mid-2026, but the cases from the 6th and 9th Circuits set the pattern courts will follow. Individual judges in the Northern District of Illinois have issued standing orders on AI use in filings. Check your assigned judge's standing orders before every AI-assisted brief.

A Verification Workflow That Holds Up

AI tools, including Claude, are genuinely useful for legal research and drafting when paired with a real verification step. Here is a practical workflow your firm can put in place this week. For a broader look at how AI agents can assist with legal work, including discovery review and conflict checks, the post on Claude Managed Agents for Wilmette law firms covers the architecture in detail.

The underlying principle is to use AI where it performs well: structuring arguments, improving prose clarity, flagging logical gaps in your theory. Do the citation work yourself in a verified legal database. If you want to understand how North Shore firms are safely adopting AI for client-facing work more broadly, the post on proactive agents for Lake Forest law firms covers the practical setup and cost tradeoffs.

When and How to Disclose AI Use

The disclosure landscape is changing faster than most practitioners track. As of mid-2026, roughly a dozen federal district courts have local rules or standing orders requiring some form of AI disclosure in filings. The Northern District of Illinois has not issued a universal rule, but individual judges in the district have issued standing orders. Before filing, check your assigned judge's standing orders on AI use. A quick search of the judge's docket for "artificial intelligence" or "generative AI" in recent orders takes ten minutes and can save significant professional consequences.

Even where disclosure is not required, consider adding a short certification to every AI-assisted brief: "Counsel used AI tools to assist in drafting portions of this brief. All legal citations were independently verified in Westlaw prior to filing." Two sentences. No filing cost. The entire risk posture of the document shifts.

The courts reviewing these cases are consistent on this point. Attorneys who disclose errors promptly and take responsibility receive reduced sanctions. Attorneys who deny AI involvement and then walk it back under pressure receive the harshest outcomes. The 9th Circuit's six-month suspension was driven as much by the concealment as by the underlying error. Candor is free. Concealment is expensive.

For a Highland Park or Winnetka firm with a federal practice, building the disclosure habit now, before you ever face a challenge, is the right approach. It also signals to clients and courts that your firm takes verification seriously, which is increasingly a real differentiator as more litigators adopt AI tools without the controls to match.

If you want to talk through building a safe AI workflow for your North Shore law firm, including verification protocols and disclosure language that fits your practice, reach out for a free 30-minute AI audit. Bace Agency works with law firms across Lake Forest, Highland Park, and the broader North Shore to build AI workflows that are both useful and defensible.

Frequently Asked Questions