Your Name Isn’t on the Brief – Until It Is

Generative AI has quietly found its place in the day-to-day operations of many personal injury firms. It helps with case research, citation checks, demand-letter drafting, motion support, and organizing case timelines. In most instances, the process appears seamless. Documents are filed, demands are sent, and cases move forward without incident.

Then a judge tries to locate four of the eight cases cited in a filing, and discovers they do not exist.

That is when the real problem begins.

How It Unraveled

A 2025 federal sanctions order offers a cautionary example of how quickly AI-assisted work can create serious consequences. In that case, a plaintiff’s response letter, signed and submitted by the supervising attorney, contained eight legal citations. When the court reviewed the filing, it could not locate four of the cited authorities. A show-cause order soon followed. The attorney explained that a paralegal had taken the initiative to assist with legal research and had used AI-based tools to generate citations. Those citations were passed along without confirming whether the cases actually existed. The attorney said she relied on the paralegal as “a diligent and trusted member” of the team and didn’t independently verify each citation before filing.

The consequences reached everyone involved. The attorney was sanctioned $1,000 under Rule 11, and the law firm was held jointly responsible . The paralegal was ordered to submit a sworn declaration to the court explaining her role in generating the fabricated citations. Both, the attorney and the paralegal filed formal apologies as part of the sanction’s proceedings. The firm was also ordered to notify the client directly.

What makes the court’s reasoning particularly significant is that it focused on a duty that already existed long before AI entered the legal profession. Rule 11 (Federal Rule of Civil Procedure 11) requires attorneys, at a minimum, to read and confirm the existence and validity of every legal authority they rely on. Because the attorney made no inquiry into the cases provided by her paralegal, the court found subjective bad faith. The paralegal’s initiative however well-intentioned was the origin of that.

In a separate case, an attorney disclosed that his paralegal drafted pleadings and briefs using AI and that the attorney would “tweak” them but not always review them carefully. Rather than reducing sanctions, that disclosure intensified the court’s response: the judge found that a paralegal drafting court submissions without meaningful attorney review constituted the unauthorized practice of law.

This Is Not an Isolated Problem

These incidents are part of a much broader trend.

As of April 2026, Damien Charlotin’s AI Hallucination Cases Tracker had documented 1,348 incidents worldwide involving AI-generated errors in legal proceedings. The pace of reported cases accelerated dramatically over a relatively short period. What began as roughly two documented incidents per week in early 2025 grew to two or three new incidents every day by the end of that year.

The tracker currently identifies 511 licensed attorneys and 33 other legal professionals, including paralegals, among the participants involved in those matters. Those numbers likely tell only part of the story.

Many state trial-court decisions never appear on Westlaw or Lexis, and countless filing errors are corrected before they become part of a published opinion. The publicly documented cases represent only the situations that made their way into the record.

For paralegals working in high-volume personal injury practices, the exposure is obvious. Demand letters, discovery responses, medical-record reviews, and motion support often move on tight timelines. In many firms, AI tools have been incorporated into those workflows informally. There may be no written procedures, no verification requirements, and no documented approval process.

That lack of structure is precisely what courts are beginning to examine.

Five Things PI Paralegals Should Do Right Now

1. Verify Every Citation Before It Leaves Your Desk: Every AI-generated citation should be independently confirmed through Westlaw, Lexis, or another reliable legal research platform. One must not assume a citation being legitimate because it looks convincing. AI-generated authorities often contain proper formatting, realistic case names, and persuasive summaries. They can still be entirely fictitious. If a citation cannot be verified, it should never be forwarded for use.

• Flag AI Use Explicitly When Handing Off To The Supervising Attorney: A simple note “citations generated with AI assistance; please verify before filing” creates a paper trail and ensures the review step is not skipped or assumed.

• Do Not Take Research Initiative Without A Documented Instruction To Do So: In the case above, the paralegal acted without being asked and the court noted this. Unsolicited AI-generated research that ends up in a filing without verification is a risk that starts at the paralegal’s desk, regardless of who signs the document.

• Ask Your Firm For A Written AI Use Policy: Who is responsible for verifying AI-generated work before it goes to court should be answered in writing, and not assumed. A verbal understanding is no protection when a show-cause order arrives.

• Understand That You May Be Drawn Into The Process: A sworn declaration to the court may be one’s outcome and not just the attorney’s. The paralegal in this case submitted a formal declaration explaining her conduct directly to federal court. The professional and reputational consequences are real, even if the monetary sanctions come out of the attorney’s account.

The Takeaway

AI can be an effective tool. It can accelerate research, organize information, and improve efficiency in busy practices. None of that changes the fundamental obligation to verify the accuracy of legal work before it reaches a court. The lesson emerging from these cases is straightforward, as responsibility does not disappear simply because a machine generated the information.

A fabricated citation can trigger sanctions. A poorly reviewed AI-generated filing can raise unauthorized-practice concerns. And when courts begin asking questions, they may look beyond the attorney’s signature and examine everyone who participated in the process.

One’s name may not appear on the brief, but if something goes wrong, it may not stay off the record for long.