Before you submit materials to a Court, confirm your citations and cases.

It is always important that you review cases you cite to make sure they say what you argue they say. And you are free to read a case and reach a conclusion that is different from someone else’s.

In the old days, cases were found in books so we had to read them. AI seems to have changed that as research is now done by algorithm which we know can make errors.

Which, arguably, makes it even more important to confirm citations, and read decisions. We have to ask why some skip this step.

A point clearly made in a recent decision of the U.S. District Court for the Northern District of Alabama. Yes, we know, a little far from home. But the issues are the same.

In Johnson v. Dunn, Case No. 2:21-cv-1701-AMM, Mr. Johnson, who was incarcerated, accused Mr. Dunn, the former Commissioner of the Alabama Department of Corrections, of fabricating case citations in two motions. Mr. Johnson was proven correct. And the Court was not pleased with Mr. Dunn, or rather his counsel.

Of the five citations in issue, one was an actual case but said nothing on the issue it was cited for, one case from 2021 was cited for a proposition but the only case with the name was from 1939 and dealt something different, one case did not exist at all, another did not exist but the Westlaw number cited did but was to a maritime personal injury case, and the last case, though the style existed, no case with the style and proposition it was cited for existed. All this could have been avoided had someone confirmed the citations and read the decisions.

Five lawyers plus the firm itself (Butler Snow) were called before the Court. Two lawyers were absolved as they had no part in the impugned materials. The firm itself was also absolved after it represented its AI policies to the Court:

…[the firm] described its artificial intelligence policies both prior to and in response to this episode…Mr. Watson stated that in June 2023, all Butler Snow attorneys received an email “stating that ‘there are significant risks that LLM [Large Language Model] output can appear perfectly researched and logical while in fact it is wholly inaccurate.”…That same email implemented a policy which requires “written permission from the appropriate Practice Group Leader to use this new technology as a secondary research tool, with full checks of the accuracy of any results through traditional legal research methods.” …Similarly, when Butler Snow provided its attorneys access to Westlaw’s CoCounsel platform in January 2025, it “adopted and distributed a policy” that “[a]ll outputs must be reviewed and verified by the responsible attorney before being presented to clients, filed with courts, or otherwise relied upon.”

Which leads to an important lesson for firms – make sure you have a policy on the use of AI tools.

That left three lawyers, one of whom was counsel to the firm, another was a partner and assistant group practice leader in the firm’s constitutional and civil rights litigation group and the third was the practice group leader of that group.  Though all three admitted their culpability, the Court was not kind to any.

The first lawyer drafted, signed and personally filed the motions that included the fabricated citations that he did not review. In discussing him the Court stated:

To be clear, not every error in a motion is recklessness or more. To err is human, and minor typographical errors, even in citations, occasionally occur despite attorneys’ best efforts. Likewise, some factual or legal authorities are the subject of reasonable debate, and a mere disagreement with one side’s view does not necessarily mean that the view is objectively false. The insertion of bogus citations is not a mere typographical error, nor the subject of reasonable debate.

The second of the three utilized AI to generate the citations but did not verify them. In discussing him the Court stated:

..[his] misconduct was more than mere recklessness. In the light of repeated general warnings from federal courts about the risks of bogus citations generated by AI, as well as the persistent specific warnings, policies, and expectations of his colleagues and law firm with respect to AI, [his] misconduct was particularly egregious. Having been so extensively alerted of the risk that AI will make things up, and having blown through all of his firm’s internal controls designed to protect court filings from counterfeit citations, [his] decisions to parrot citations generated by AI without verifying even one of them reflect complete and utter disregard for his professional duty of candor. This is recklessness in the extreme, and it is tantamount to bad faith…

To the third of the three, the most senior, the Court had some more pointed comments:

On this factual record, the court has no difficulty finding that [he] bears responsibility for the false statements of law made to the court over his name in the signature block….Indeed, although [he] did not personally use AI to generate citations and did not personally file the motions at issue, the record does not suggest that he would have done anything differently…nor that he expected [others] to do anything differently. According to [his] own testimony, he did not make any effort to verify the contents of the motion to compel before authorizing its filing, and it would have been extremely unusual for him to do so. Nor did he require (or even ask)…any other attorney (or person), to undertake that task. Nor was it his practice to require (or ask) that of them.

…[He] simply assumed the truth of what was in the draft, and/or assumed that someone else would check on that. This is the same indifference to the truth and complete personal disinterest in the most basic professional responsibility…. Particularly in the light of [his] roles as practice group leader, supervisory attorney, and partner — and the reality that he is the only lawyer on the team entrusted with the necessary deputy attorney general designation — this utter disregard for the truth of filings bearing his name in the signature block is particularly egregious, more than mere recklessness and tantamount to bad faith.

To be clear, the court’s finding in this regard is not simply a harsh inference: when it became apparent that multiple motions with his name in the signature block contained fabricated citations, [his] nearly immediate response was to try to skip the show cause hearing and leave the mess for someone else. And when the court compelled him to appear at the hearing, he paired his apology with an explanation in greater fullness of how very little work he personally puts in to be sure that his team’s motions tell the truth. This cannot be how litigators, particularly seasoned ones, practice in federal court or run their teams…

The Court’s sanction order for the three included public reprimands, the requirement that they provide a copy of the Order to all their clients, that the Order be published in the Federal Supplement and referral to the State bar.

Feel free to use AI research tools. But going back to something from the early days of law school – caveat emptor – read and confirm the output.