Medical Buyers Group v. Pence

What sanction did the court impose?

Judge Tilman E. Self, III publicly admonished Gordon in a November 18, 2025 order (ECF No. 52) for submitting hallucinated and false case citations, and invited defendants to seek their reasonable attorney's fees. On January 20, 2026 (ECF No. 66), the Court awarded $10,000 in sanctions ($5,000 to each defense team) under its inherent authority. The Court explicitly held it could not act under Rule 11 because defendants had not filed separate sanctions motions and had not afforded the 21-day safe harbor required by Rule 11(c)(2), and could not act sua sponte under Rule 11(c)(3) because defendants had requested sanctions in their replies. Citing In re Mroz, 65 F.3d 1567 (11th Cir. 1995), the Court invoked its inherent authority instead, holding (in line with U.S. v. McGee, 2025 WL 2888065 (S.D. Ala. Oct. 10, 2025), Johnson v. Dunn, 762 F. Supp. 3d 1241 (N.D. Ala. 2025), and Benjamin v. Costco Wholesale Corp., 779 F. Supp. 3d 341 (E.D.N.Y. 2025)) that even unintentionally citing AI-hallucinated cases constitutes bad faith. The fee award was reduced from defendants' combined $21,589.33 request: the Court declined to require Gordon to pay for the show-cause hearing attendance the Court itself had compelled, applied a compensatory rather than punitive standard under Goodyear Tire & Rubber Co. v. Haeger, 581 U.S. 101 (2017), and aimed at rough justice rather than line-item auditing.

Why does Medical Buyers Group v. Pence matter for law firms using AI?

The procedural twist is the load-bearing point for firm readers. Defendants flagged Gordon’s fake citations in their reply briefs and asked the Court for sanctions. That request, the Court held, defeated both Rule 11 paths: there was no separate Rule 11 motion and therefore no 21-day safe harbor under Rule 11(c)(2), and the Court could not act sua sponte under Rule 11(c)(3) because defendants had already requested the sanctions. A line-by-line read of Rule 11 would end the analysis there. The Court did not stop, citing In re Mroz, 65 F.3d 1567 (11th Cir. 1995) for the proposition that the Court’s inherent authority to sanction bad-faith conduct survives Rule 11’s procedural limits.

The bad-faith finding rested on accumulated 2025 federal precedent rather than anything specific to Gordon’s intent. The Court quoted U.S. v. McGee for the line “Given the broad attention given to the problem with AI generated cases, no attorney can claim ignorance,” and cross-referenced Johnson v. Dunn (N.D. Ala.) and Benjamin v. Costco (E.D.N.Y.) for parallel holdings that sincere apologies and remedial CLE work do not disturb the bad-faith characterization. For partners reading this case as precedent, the practical effect is that the standard of care is now calibrated to constructive knowledge of AI hallucination risk: firms that want to brief AI-assisted work without a Rule 11 fight need a written verification protocol and supervisory record, not a “we did not know” defense.

The fee award itself reflects measured judicial rough justice rather than the full $21,589.33 defendants sought. Judge Self took out the time defense counsel spent attending the show-cause hearing the Court had ordered them to attend, declining to make Gordon pay for the Court’s own procedural choices, and applied the Goodyear compensatory standard to cap the award at an amount the Court considered reasonable for the actual research-and-response burden imposed by the fake citations. The result is a useful template: $5,000 per defense team is the realistic floor when a sole-practitioner-pattern AI-hallucination defendant takes responsibility quickly and does substantial remedial work before the fee hearing.