Williams v. Dallas-Fort Worth International Airport Board

What did the court warn about?

District Judge Mark T. Pittman granted the motion to dismiss the Section 1983 claims, holding that plaintiffs identified no clearly established law defeating qualified immunity and did not plausibly allege municipal liability, and declined supplemental jurisdiction over the state-law claims. It dismissed the Section 1983 claims with prejudice and the state-law claims without prejudice "so they can be pursued in Texas state court." On the citations, the court stated that "Plaintiffs make a borderline sanctionable attempt to identify 'clearly established' law from five cases" and that it "has little patience for such 'lazy lawyering.'" It imposed no sanction and issued no show-cause order. In footnote 2 the court cautioned "that use of generative artificial intelligence that results in repeated misrepresentations of law may qualify for sanctions under Fed. R. Civ. P. 11(b)" and (c), citing Texas Disciplinary Rules of Professional Conduct 3.01, 3.03(a)(1) and 8.04(a)(3), and noted that the Fifth Circuit "recently approved a $2,500 sanction for AI hallucinated misrepresentations" in Fletcher v. Experian Info. Sols., Inc., 168 F.4th 231 (2026). The order does not name plaintiffs' counsel.

Why does Williams v. Dallas-Fort Worth International Airport Board matter for law firms using AI?

Williams is a qualified-immunity dismissal in which defective citations were part of why the plaintiffs’ argument failed on the merits, not a separate sanctions proceeding. The clearly-established-law inquiry, as the court framed it, quoting the Fifth Circuit in Joseph v. Bartlett, requires a case “in which an officer acting under similar circumstances was held to have violated the [Constitution],” and “it certainly cannot be established by citation to nonexistent cases.” Of the five cases plaintiffs offered, the court said that for Dawkins it “cannot even find a case purporting to ‘clearly establish’ law on the matter,” said plaintiffs “presumably meant” a 2008 Fourth Circuit decision when they cited Iko, and found Yates both out-of-circuit and factually distinct. The remaining two were real Supreme Court decisions: Garner, which the court held plaintiffs had read too broadly, and Graham, which it held did not help them.

The court’s response to the citations was verbal rather than remedial. It called the attempt “borderline sanctionable” and “lazy lawyering,” but issued no show-cause order and imposed no sanction. The AI reference appears only in footnote 2, which cautions that generative AI use resulting in “repeated misrepresentations of law may qualify for sanctions” under Rule 11 and the Texas disciplinary rules, and notes that the Fifth Circuit “recently approved a $2,500 sanction for AI hallucinated misrepresentations” in Fletcher v. Experian. The order does not say that plaintiffs’ counsel used AI.

Firms documenting compliance may wish to consider that in this case the cost of unverified authority was borne first by the client: the court held that the cases offered to meet plaintiffs’ burden did not carry it, and it put counsel on notice that a repetition could be treated as sanctionable.