Leslie v. IQ Data Int'l

What sanction did the court impose?

Magistrate Judge Regina D. Cannon issued a Final Report and Recommendation (ECF No. 98) on June 5, 2024 recommending dismissal with prejudice as a sanction under Rule 37(b)(2). The R&R found Leslie's discovery non-compliance willful and his citation to nonexistent case law 'further evidence of Plaintiff's bad faith,' citing Thomas v. Pangburn, 2023 WL 9425765 (S.D. Ga. Oct. 6, 2023) and Willy v. Coastal Corp., 855 F.2d 1160 (5th Cir. 1988) for the proposition that fake citations alone are sanctionable. The R&R declined to also recommend monetary sanctions, finding Leslie's pro se circumstances would make a fee award unjust under Fed. R. Civ. P. 37(b)(2)(C). On November 25, 2024 (ECF No. 108), District Judge Victoria M. Calvert adopted the R&R in full, dismissed the case with prejudice, and held that Leslie's 'repeated failure to produce the documents at issue over the span of 13 months, coupled with his repeated citation of nonexistent case law, evinces Leslie's bad faith and justifies this significant sanction.'

Why does Leslie v. IQ Data Int'l matter for law firms using AI?

The Eleventh Circuit framework for sanctioning fake citations does not require an attorney target or an AI admission. Magistrate Judge Cannon’s R&R is structured to make that point: the discovery non-compliance over 13 months establishes willfulness under Goodman v. New Horizons, the citation to nonexistent authority adds a second independent basis for bad faith under Thomas v. Pangburn and Willy v. Coastal Corp., and the combined record supports dismissal with prejudice as the only sanction that would deter further conduct. None of that analysis turns on whether Leslie used ChatGPT or another tool, and the order is careful not to assume.

The case is most useful as a template for fake-citation procedure rather than as an AI-policy precedent. The defense team made the bad-faith finding possible by docketing the fake-citation challenge in its reply brief, then following up with a notice when the misconduct repeated in the sur-reply. The Court could then ask Leslie about the citations on the record at the teleconference, and his non-answer (‘good case law,’ nothing more) became the basis for the bad-faith finding under both Rule 37 and Rule 11. For partners who practice against pro se litigants increasingly using generative AI tools without disclosure, the procedural moves here are reusable: flag the fake cite in writing on the docket, request a teleconference where the citing party has to explain, and let the absence of explanation do the bad-faith work.

The R&R also reflects an emerging pattern in pro se sanctions cases of declining monetary fees in favor of dismissal. Cannon’s R&R notes Leslie’s circumstances and concludes that a fee award would be unjust under Rule 37(b)(2)(C), citing the Court’s discretion. District Judge Calvert’s adoption order reinforces the same calibration. For firms tracking the cost-of-sanctions question across the AI-hallucination case set, this is the pro se analogue to the Johnson v. Dunn approach in attorney cases: monetary sanctions are not where these courts are landing when the misconduct is severe enough to support dismissal or disqualification.