Five sanctions from the last two weeks

This one didn't come out of ChatGPT — the record names Westlaw Precision, a paid legal-research product, as the tool behind the fabrications. The motion cited real, correctly-named cases (Grubbs v. Norris, In re Dinnubilo, In re Symington, Wilcher, Bennett Funding Group) and then attached quotations and holdings to them that none of those opinions actually contain: a privilege rule Grubbs doesn't support, a "premature and unauthorized substitute for formal discovery" line invented and pinned on Dinnubilo, a Rule 2004 limitation attributed to Symington that isn't there.

That's five separate fabricated or misrepresented authorities in one motion, and the court didn't stop at striking them. It held counsel in civil contempt, ordered $29,877 in adverse costs, and mandated continuing legal education specifically on generative AI — a financial, professional, and remedial penalty stacked in a single order.

The detail worth sitting with: every case name and citation was real. The fabrication was entirely inside the quotation marks.

A pro se litigant's brief cited "Barton v. McGovern, 302 So. 3d 1042" — a case the court found doesn't exist anywhere in the Florida reporters. Separately, the same brief quoted language and claimed it came from an actual Barton decision (504 So. 2d 457, a real 1987 case) that never contains the quoted passage. Two different failure modes wearing the same case name.

The brief didn't stop there — the court's order describes "multiple other phantom authorities and fabricated quotations" scattered through the filing, including an invented quote pinned on Stockman v. Downs.

For a lawyer, a bad citation risks a sanction. For someone representing themselves, the Fourth DCA's response was to open a proceeding to consider barring the litigant from filing anything further pro se in that court — the sanction that ends the case, not just this brief.

This is a subtler fabrication pattern than an invented reporter citation: the court found that four separate citations pointed to real, existing case reports — just not the cases the brief claimed were there. The correct case at one cited citation is MVA v. Deering, 438 Md. 611; at another it's Johnson v. Mayor & City Council of Balt., 387 Md. 1; at a third it's Marshall v. Holmes, 141 U.S. 589. The citation numbers resolved. The case names attached to them didn't.

On top of the four mislabeled citations, the court also found two instances of quoted language that doesn't appear in the actual opinions, plus one authority cited for a proposition it doesn't support — six distinct citation defects across a single appellant's brief.

The court's response was to strike the affected portions of the brief rather than let the appeal proceed on citations that didn't check out.

In an opposition to PG&E's motion, the plaintiff cited two California appellate cases — "Boling v. Pub. Utils. Comm'n, 105 Cal.App.3d 805" and "Loving v. Cnty. of Stanislaus, 33 Cal.App.5th 444" — that the court found simply don't exist. A third citation, Pierce v. Cnty. of Orange (a real Ninth Circuit case), was real but mischaracterized: the court noted the opinion exists but doesn't address the statute-of-limitations or continuing-violation argument it was cited to support.

Three defects in close proximity — two phantom cases and one real case doing work it was never cited for — is the density that tends to make a court stop trusting an entire section of a brief once the first one is caught.

The court's response here was measured — admonishment and a warning, not fees or contempt — but a warning on the docket in active utility litigation is a credibility mark that follows counsel into every filing that comes after it.

The plaintiff cited two cases using realistic Westlaw-style reporter numbers — McCarty v. United States at 2015 WL 1543969, and Harris v. Dallas Cnty. Hosp. Dist. at 2022 WL 17830420. The format alone made both look like ordinary case citations. The court couldn't locate either one, and when asked to produce a PDF of the underlying opinion, the plaintiff couldn't.

The resulting monetary sanction — $610 — is modest next to the six-figure orders elsewhere in this database, but the court's process didn't scale down with the dollar amount: cite something that can't be verified, and get asked to prove it exists.

What makes this one worth including isn't the size of the sanction, it's the fabrication style. A WL-format citation number is specifically the kind of thing that reads as legitimate on a skim, because it looks exactly like every other Westlaw citation in the brief.