Whiting v. City of Athens

In Whiting v. City of Athens (March 2026), the Sixth Circuit imposed $15,000 sanctions on each of two attorneys plus full payment of opposing counsel's fees for filing a brief containing 24 fabricated case citations. The citations followed a pattern consistent with AI hallucinations — plausible case names, correct reporter formats, real court names, but entirely fictional decisions. One fabricated case, "Berg v. Knox Cnty.," was cited six times across two briefs.

This is the most significant appellate sanctions ruling of 2026 and the first time a federal circuit court has imposed five-figure individual sanctions for AI-generated hallucinated citations. The Sixth Circuit's opinion makes clear that "I didn't know the AI made them up" is not a defense — attorneys have an independent duty to verify every citation they file, regardless of how it was generated.

The 24 Fabricated Citations and How They Were Caught

The two attorneys filed an appellate brief citing 24 cases that don't exist. The fabricated citations were scattered across the statement of facts, argument section, and reply brief — not concentrated in one area, which suggests multiple rounds of AI-generated content were pasted into the filings without verification.

Opposing counsel flagged the problem after failing to locate "Berg v. Knox Cnty." — cited six separate times as controlling Sixth Circuit precedent. A systematic check revealed 23 additional phantom cases. Every fabricated citation followed the same pattern: real-sounding party names, correct citation formats (F.3d, F. Supp. 2d), actual court names, but zero matches in Westlaw, Lexis, or any other legal database.

The court noted that this pattern is "consistent with the known output characteristics of large language models when generating legal citations" — the first time a federal appeals court has explicitly identified AI hallucination patterns in a published opinion.

The Sixth Circuit's Sanctions Analysis

The panel didn't mince words. The court found that the attorneys violated Federal Rule of Appellate Procedure 46(c) and their obligations under Rule 11 (incorporated by reference). The sanctions broke down into two components: $15,000 per attorney as a direct penalty, plus full reimbursement of opposing counsel's attorney fees for the time spent identifying and briefing the fabricated citations.

The court rejected every defense the attorneys raised. "Inadvertence" didn't work — 24 fake citations across multiple filings isn't a typo. "Reliance on research tools" didn't work — the court held that attorneys can't outsource verification to any tool, whether it's a junior associate, a research service, or an AI model. "No client harm" didn't work — the court found that filing fabricated authorities wastes judicial resources and undermines the integrity of the appellate process.

The opinion explicitly states: "The duty to verify citations is non-delegable. It belongs to the attorney whose name appears on the filing."

Why This Case Is Different from Prior AI Sanctions Cases

Previous AI citation sanctions — including the widely reported *Mata v. Avianca* case (S.D.N.Y. 2023) where attorneys were fined $5,000 — were district court rulings with limited precedential weight. Whiting is a published Sixth Circuit opinion, binding on all federal courts in Ohio, Michigan, Kentucky, and Tennessee.

The sanctions amount is also significantly higher. Prior cases imposed $1,000-$5,000 penalties that critics called insufficient to deter AI misuse. The Sixth Circuit's $15,000-per-attorney penalty plus full fee-shifting signals that appellate courts will treat fabricated citations as serious misconduct, not a minor embarrassment.

The opinion also goes further than prior cases in articulating why AI-generated citations are particularly dangerous: they're designed to look plausible, they follow correct formatting conventions, and they exploit attorneys' tendency to skim-verify citations that "look right." The court called this "a systematic risk that demands systematic verification."

The Verification Standard Going Forward

The Sixth Circuit laid out what adequate verification looks like: every case cited must be independently confirmed in a primary legal database before filing. Not spot-checked. Not verified by the same AI that generated the citation. Every single one.

The court acknowledged this standard increases the cost and time of brief-writing. It didn't care. "The integrity of citations to legal authority is foundational to the adversarial system," the opinion states. "If AI tools reduce the cost of generating arguments, some of that savings must be reinvested in verification."

For firms using AI in appellate practice, the takeaway is operational. You need a verification step built into every brief-writing workflow — a human attorney who checks each citation against Westlaw or Lexis before the brief goes out. This should be documented in the file. If sanctions are ever at issue, you want a paper trail showing independent verification.

Implications for Law Firm AI Policies

Whiting makes law firm AI policies a risk management necessity, not a nice-to-have. Any firm that allows attorneys to use AI for research or drafting without a mandatory verification protocol is accepting the risk of five-figure sanctions and fee-shifting — per attorney, per filing.

The practical requirements are straightforward. First, mandatory citation verification against a primary legal database before any filing. Second, documentation that verification was performed and by whom. Third, training on AI hallucination patterns so attorneys know what to look for. Fourth, supervision protocols requiring a senior attorney to review AI-assisted filings.

Firms that had informal AI policies before Whiting need to formalize them now. The Sixth Circuit's published opinion means that in any future sanctions proceeding, the court will ask whether the firm had a policy — and whether the attorney followed it.

The Bottom Line: The Sixth Circuit's $15,000-per-attorney sanctions in Whiting establish the appellate standard: every AI-generated citation must be independently verified, and "I didn't know" isn't a defense.