Fined and suspended: when a court decides money isn't deterring fake citations

July 11, 2026 · Every factual claim in this piece links to its source.

The going rate for a first fabricated-citation sanction has been a few thousand dollars and an apology. In Twigg v. BSN Sports, Inc., No. 4:23-CV-00067-MWB (M.D. Pa.), Chief Judge Matthew W. Brann reached for something that costs an attorney more than money. As reported by Daniel Cummins at Tort Talk and by Riley Brennan in The Legal Intelligencer, the June 18, 2026 sanctions order fined the attorney $1,500 and suspended him from practicing in the Middle District of Pennsylvania for six months for submitting filings containing AI-generated hallucinated citations.

The dollar figure is small. The suspension is the point. A six-month bar from a federal district is not a line item — it is time a lawyer cannot appear before that court, and it signals a court concluding that fines alone have stopped working. The Sixth Circuit said the quiet part out loud in Whiting three months earlier: "smaller fines have plainly been inadequate." Twigg is what the next step looks like.

Two kinds of bad citation, one rule

The court found the attorney violated Rule 11 by submitting "fabricated and inaccurate case citations," per Tort Talk's summary of the opinion. Those are two different failures, and the distinction is the whole game:

Fabricated — a citation to a case that does not exist. The reporter, the volume, the page, the parties: invented. This is the failure a machine is actually good at catching, because "does this case exist in the public record?" is a lookup, not a judgment call.
Inaccurate — a citation to a real case that doesn't say what the brief claims, or a real reporter cite attached to the wrong case name. Some of these a lookup catches (the name and the citation don't match the same record); the ones that turn on what a real opinion holds require a human to open the case and read it.

A generative model produces both kinds fluently, and it produces them in the same confident register as correct ones. That is exactly why they slip into a signed filing.

The part that drew the judge's ire: shifting the blame

According to Tort Talk, Judge Brann "also expressed his displeasure with the attorney attempting to shift the blame relative to his responsibility for submitting the erroneous citations." That sentence is the one to sit with, because it is where sanctions law and AI tooling collide.

Rule 11(b) makes the certification personal: by signing and filing, an attorney certifies that to the best of their knowledge, formed after a reasonable inquiry, the filing's legal contentions are warranted. There is no carve-out for "my research tool generated it." The obligation to conduct a reasonable inquiry is not delegable to software — a point the Ninth Circuit made precedential in Lnu v. Blanche when it held that the rules are violated at signing, not at research. Blaming the tool doesn't reduce the sanction; courts have repeatedly read it as an aggravator, a failure to take responsibility.

The pattern, not the outlier: Twigg is one of a fast-growing line of 2026 decisions — the Fifth Circuit's $2,500 sanction in Fletcher v. Experian, the Sixth Circuit's removal of counsel in United States v. Farris, and others catalogued in the AI Hallucination Cases database — sourced here from Norton Rose Fulbright's 2026 sanctions survey. The through-line: the citation was never verified before it was filed.

What actually prevents the Twigg outcome

Not a policy memo. A step in the workflow, run every time, before the signature goes on the filing. The reasonable inquiry Rule 11 demands, for citations specifically, has a concrete floor: confirm each cited case exists in the public record, confirm the citation matches the case name it's attached to, and read anything you're relying on for its holding. The first two are mechanical and are exactly what get skipped under deadline. The third is judgment and can't be automated — but it's a lot faster once the fabricated and mismatched cites have already been pulled from the pile.

Where an automated check helps — and where it stops

Fabricated cites — caught. A citation to a case with no counterpart in the public record fails the lookup. That's the "does this case exist?" question answered before you rely on it.
Name/citation mismatches — caught. When a real reporter cite is attached to a case name that doesn't correspond to it, a name-to-citation match flags the disagreement — the mismatch that survives a quick "did something pull up?" glance.
Misdescribed holdings — not caught, and no honest tool claims otherwise. A real case cited for a proposition it rejects lives inside a real opinion. Catching it means reading the opinion. An existence-and-match check clears the board so your reading time goes to the cites that need it.
Westlaw/Lexis-only cites — reported as unverifiable, not fabricated. Database numbers (e.g., a "2026 WL …" cite) aren't in public court records, so LegalCite returns unverifiable rather than guessing — which is precisely the cite to open and read yourself before filing.

Before you sign it, run it. LegalCite checks every citation in a brief against public court records, flags cases that don't exist, flags citations that don't match their case names, and honestly reports Westlaw/Lexis database cites as unverifiable rather than guessing. Free, no signup, and the document never leaves your browser: run it through Brief Check. It won't read the holdings for you — nothing can — but it makes the reasonable inquiry the record shows these lawyers skipped a two-minute step instead of a good intention.

LegalCite verifies that citations exist in public court records (CourtListener) and match the case names attributed to them. It does not assess good-law status or verify quotations, holdings, or record accuracy — those require human review — and nothing here is legal advice. Citations to unpublished databases (Westlaw/Lexis) are reported as unverifiable, not fabricated. Case details reported second-hand from the sources below; readers relying on Twigg should consult the docket and orders directly. Sources: Tort Talk (Daniel Cummins), "Artificial Citations Lands Another Attorney in Trouble in Pennsylvania," July 2026, summarizing Twigg v. BSN Sports, Inc., No. 4:23-CV-00067-MWB (M.D. Pa. June 18, 2026 sanctions order; underlying opinion May 21, 2026, Brann, C.J.); Riley Brennan, "Fed. Court Suspends Pa. Attorney for AI-Hallucinated Citations," The Legal Intelligencer, June 22, 2026; Fed. R. Civ. P. 11 (Cornell LII); Norton Rose Fulbright, "AI in litigation: Update on Gen AI sanctions in 2026"; AI Hallucination Cases database (Damien Charlotin).