By Vinitha Prasannan, Senior Manager – Delivery
Fake AI Citations Got a Deutsche Bank Brief Thrown Out and Exposed a Blind Spot in Court Rules
Courts have repeatedly warned lawyers about fabricated authorities produced by generative AI. The recent case law, Douglas v. Deutsche Bank National Trust Co., No. 24-CV-1099 (D.C. Ct. App. Sept. 3, 2026), however, raises a more unusual question.
What happens when a court identifies serious AI-related misconduct, but its existing rules do not clearly authorize a direct and proportionate sanction against counsel?
That Douglas issue is distinct beyond the familiar “verify your citations” discussion. The D.C. Court of Appeals struck the appellee’s brief after finding fabricated authorities, but Senior Judge Stephen Glickman’s concurrence identified a broader procedural problem. The court’s sanctioning authority may not fully reach this type of attorney misconduct.
How Four Made-Up Citations Turned a Simple Appeal Into a Legal Mess
Douglas arose from a judicial foreclosure appeal. While reviewing Deutsche Bank’s appellee brief, the D.C. Court of Appeals discovered several authorities it could not locate or verify.
The court issued a show-cause order asking why the brief should not be stricken for citing non-existent cases that appeared to result from AI hallucinations. Counsel later acknowledged that four authorities did not exist and explained that she had used Google’s generative-AI search tool to locate case authority without independently verifying the citations before filing.
The court ultimately struck the brief and referred the matter to the Office of Disciplinary Counsel.
But the more significant issue appeared in Judge Glickman’s concurrence.
The Court Could Strike the Brief, But Its Power to Sanction Counsel Was Less Clear
Judge Glickman described Douglas as the first reported case in which the D.C. Court of Appeals had received a brief containing fabricated citations. He agreed that striking the brief under D.C. Appellate Rule 28 was appropriate, but called that remedy a “comparatively weak response” because it also penalized the innocent client for counsel’s conduct.
Federal courts confronting similar conduct have imposed more targeted sanctions, including monetary penalties, attorney-fee awards, disciplinary referrals, disqualification, notice requirements, and suspension. Judge Glickman stated that such sanctions should be available to punish and deter fabricated citations. Yet he concluded that the D.C. Court of Appeals’ authority to sanction counsel directly was “more limited than one might have supposed.”
That distinction gives Douglas its strongest legal significance.
The problem was not whether the conduct warranted consequences. The problem was whether the court had the right procedural method to impose them.
Rule 28 Allowed the Court to Strike the Brief
The panel relied on D.C. Appellate Rule 28, which requires appellate briefs to contain a party’s contentions and reasons supported by citations to authorities and the record.
The court reasoned that a fabricated citation is “at best, no citation at all; indeed, it is worse.” Based on that noncompliance, the panel struck the appellee’s brief.
Rule 28, however, primarily governs the content of appellate briefs. It does not provide the same range of attorney-specific disciplinary remedies available under other rules.
That limitation became clearer when the concurrence considered Rules 38 and 46.
Rule 38 Does Not Cleanly Reach a Brief Containing Fake Citations
D.C. Appellate Rule 38 authorizes sanctions when a party or attorney files a frivolous appeal, petition, or motion, or acts for an improper purpose. The rule does not expressly identify briefs.
Judge Glickman also explained that D.C. precedent sets a demanding standard for frivolousness. A filing must essentially lack any meaningful legal substance or realistic chance of success.
That creates an important distinction. A brief can contain fabricated authorities without being entirely frivolous if legitimate authority independently supports the underlying legal argument. In that circumstance, the citation misconduct remains serious, but Rule 38 may not fit.
For that reason, Judge Glickman concluded that Rule 38 could not serve as the “main, or even the usual,” basis for sanctioning briefs containing AI hallucinations.
D.C. Rule 46 Is Narrower Than the Federal Rule
The difference between the D.C. and federal versions of Rule 46 proved even more significant.
Federal Rule of Appellate Procedure 46 allows a federal appellate court to discipline an attorney for “conduct unbecoming a member of the bar” or for failure to comply with a court rule. Federal appellate courts have relied on that provision in AI-related sanctions cases.
D.C. Appellate Rule 46 does not contain equivalent language. It addresses admission to the bar and does not expressly authorize discipline for misconduct or violations of appellate rules.
As a result, one of the most obvious sanctioning tools available to federal appellate courts was not clearly available to the D.C. Court of Appeals.
Inherent Authority Also Had a Limitation – Bad Faith
The court also considered its inherent authority to sanction misconduct.
D.C. precedent requires a finding of bad faith before inherent-power sanctions may be imposed. Judge Glickman emphasized that bad faith must remain distinct from negligence, professional incompetence, or even serious mistakes.
On the existing record, he did not conclude that counsel or the firm had acted intentionally, knowingly, or even recklessly. Instead, he characterized the conduct as potentially negligent, grossly negligent, or professionally incompetent.
That created the gap at the center of the concurrence.
A lawyer may submit fabricated authority through serious negligence. The conduct may waste judicial resources and undermine confidence in the adversarial process. Yet if the record does not support bad faith, inherent-power sanctions may remain unavailable.
Judge Glickman therefore concluded that, beyond public admonishment and striking the brief, the court’s existing rules and case law appeared to prevent additional direct sanctions against the attorneys.
The Court Sent the Issue to Its Rules Committee
The panel did not ignore the structural problem.
In its sanctions discussion, the court expressly stated that it was referring the matter to its Rules Committee for analysis and possible clarification of the full scope of its sanctions authority in these circumstances.
That referral may ultimately prove more important than the sanction imposed in Douglas itself.
The case suggests that courts may need to examine whether existing procedural rules adequately distinguish among negligent AI-related errors, reckless conduct, intentional fabrication, supervisory failures, and frivolous litigation.
Those categories are not necessarily treated the same under current sanctions law.
A Secondary Issue is that Who Bears Responsibility for a Defective Brief?
Judge Glickman also addressed whether every attorney whose name appears on a brief should bear the same responsibility for fabricated citations.
He agreed with the general principle that attorneys associated with a brief bear responsibility, but rejected the idea that every listed attorney must personally check every citation. He noted that legal briefing often involves multiple lawyers with different roles and that duplicative citation review by every attorney would be impractical.
Instead, he suggested that firms can use structured safeguards, including training, clear AI-use policies, compliance procedures, and designated citation-review processes.
This shifts the focus from simply asking who used AI to asking what responsibility each lawyer had and what review system the firm maintained.
Why Douglas Matters
The most important point in Douglas is not that AI can generate fake cases. Courts have already addressed that problem repeatedly.
The more significant development is that AI-related misconduct can expose weaknesses in procedural rules drafted long before generative AI became part of ordinary legal practice.
The D.C. Court of Appeals could strike the brief. It could publicly criticize counsel. It could refer the matter for disciplinary review.
But where the conduct appeared negligent rather than intentionally deceptive, the court questioned whether its existing rules gave it authority to impose a more targeted sanction directly on the responsible lawyers.
That is what makes Douglas worth watching.AI may not be changing lawyers’ core duties of competence and candor. But it may be forcing courts to reconsider whether their existing sanctions rules are equipped to enforce those duties in a proport
For more practical litigation drafting tips, pleading and legal research insights, follow LawCompany.