A partner at a Virginia law firm has publicly eaten crow in a San Francisco federal courtroom after an AI-generated draft slipped fake quotations into a filing in the litigation over the Trump administration's mass federal layoffs. The bogus language surfaced in a motion over a subpoena dispute, forcing the lawyers to alert the judge and opposing counsel and providing fresh fuel for concerns about legal briefs drafted with generative AI but not carefully reviewed by humans.
The apology followed attorney Jason Greaves of Binnall Law Group admitting he had used Anthropic's Claude Console to crank out an initial draft of a May 6 motion to quash a subpoena, according to Reuters. Greaves told U.S. District Judge Susan Illston he sent the AI-produced draft to an associate for review but did not personally re-verify the quoted language before it went out the door.
In a subsequent filing, Greaves acknowledged, "While the cases cited may have generally supported our position, I acknowledge the inexcusable error of allowing the inclusion of phantom quotations," according to court papers cited by Bloomberg Law. The motion was tied to efforts to block testimony from Joseph Guy, a former Department of Homeland Security deputy chief of staff, in the broader AFGE v. Trump litigation.
Firm founder Jesse Binnall did not sugarcoat the misstep, calling the AI-linked errors unacceptable, inexcusable, and an embarrassment to this firm. Binnall Law Group said it will add supplemental training and require its attorneys to rely on traditional research tools such as Westlaw or LexisNexis, according to Decrypt. The filings say the firm plans additional safeguards so that anything drafted with AI is thoroughly checked before it hits the docket.
A growing pattern in courtrooms
Judges around the country have been flagging fabricated citations and mangled quotations as lawyers lean on AI tools, leading to sanctions, public scoldings and new local rules in some courts. Legal ethics experts point out that Rule 11 and long-standing professional duties place the onus squarely on the lawyer who signs a brief, regardless of how it was drafted. Analyses warn that courts are seeing a sharp uptick in AI-related filing errors, according to the UIC Law Library.
Case details and next steps
The challenged motion sought to quash a subpoena for the deposition of Joseph Guy in AFGE v. Trump, No. 3:25-cv-03698 (N.D. Cal.), according to the Civil Rights Litigation Clearinghouse. As the case moves forward, Judge Illston could order corrective filings, consider fee awards or other sanctions, or require a fully corrected brief.
Ethics commentators note that the core rule has not changed: the lawyer whose name is on the filing is responsible for every citation in it, regardless of whether AI helped with the first draft. Failing to verify authorities can still bring monetary sanctions or even disciplinary referrals. Binnall Law Group says it will roll out additional training and tighter review protocols, and the Northern District of California's response on any sanctions or remedial steps will be closely watched by firms weighing AI tools against professional risk. For now, the docket and any follow-up motions will determine whether this episode remains a one-off mea culpa or becomes another data point in the growing file of AI-related courtroom mishaps, as reflected in the Civil Rights Litigation Clearinghouse.
