Kingsfield / AI hallucination cases / Jackie L. Miller v. Regions Bank
VERDICT: REJECT

Jackie L. Miller v. Regions Bank

ChatGPT, in a suit against Regions Bank. A real case, an invented quote.

Court
N.D. Alabama
Decision date
2026-05-21
AI tool (per record)
ChatGPT
Hallucination
Fabricated quote
Court outcome
DQ from case and court (6 months); Bar Referral
Kingsfield verdict
REJECT
In tested set
Yes

What happened

Jackie Miller sued Regions Bank, and her attorney, H. Gregory Harp, filed a response opposing the bank's motion for summary judgment. The brief included four false quotations attributed to real cases, including Batson v. Salvation Army and Holly v. Clairson Industries.

The Court searched for each quotation and could not find it in the cited opinions or anywhere else. After the Court ordered Harp to produce his ChatGPT history, he deleted his ChatGPT account, then told the Court the account could not be recovered. Emails the Court later obtained showed he canceled the account three days after the production order and took a refund to cut off his access early.

The fabricated citation

Batson v. Salvation Army

Offered for: an FMLA adverse-action standard

What the court did

The Court found that Harp made false statements of law in violation of Rule 11 and publicly reprimanded him, with a publication requirement. For deleting his account, it found him in civil contempt, drew an adverse inference that ChatGPT generated the fabricated quotations, disqualified him from the case, suspended him from practice in the Northern District of Alabama for six months, and referred him to the Alabama State Bar.

How Kingsfield ruled

Kingsfield returned REJECT. The quotation attributed to this case is fabricated. The court could not find the quoted language anywhere in the cited Eleventh Circuit opinion.

This matter is among the US case-law cases we ran Kingsfield against in the Charlotin benchmark. Kingsfield's existence check confirms every citation against the case-law corpus and an independent source (CourtListener) before any claim resting on it can stand. Across the run Kingsfield ruled 696 of 700 testable claims correctly, including every one of the 504 fabricated citations and 92 fabricated quotes.

Honest scope. This measures Kingsfield's ruling layer given an extracted citation, not an end-to-end run over a full brief. The case record and the AI-tool attribution come from Damien Charlotin's public database and the court order; the verdict is Kingsfield's. See the full 700-claim benchmark.

← Back to the live verdict demo