Jones v. Department of the Army: Pro Se Litigant’s AI-Fabricated Citations Draw Federal Circuit Warning | Advocate Prakhar

Case at a Glance

Full Citation	Jones v. Department of the Army, Nos. 2025-1822, 2025-1823 (Fed. Cir. May 13, 2026)
Court	U.S. Court of Appeals for the Federal Circuit
Date	May 13, 2026
AI Tool	AI tool (unspecified); citations generated by artificial intelligence
Outcome	Affirmed-in-part, vacated-in-part, remanded; formal warning issued on AI citations; sanctions threatened for repetition

Background

Anthony Jones was appointed by the Department of the Army as an IT Specialist at a regional cyber center in Germany in 2013. He was removed from his position on July 1, 2021. He challenged his removal before the Merit Systems Protection Board (MSPB), and an Administrative Judge found the removal was the result of unlawful whistleblower retaliation — ordering the agency to cancel the removal and pay back pay with interest.

The agency failed to comply, leading Jones to file a petition for enforcement. Concurrently, he sought substantial compensatory and consequential damages — including $9.9 million in damages and punitive damages — arguing lost future wages and emotional distress. The Administrative Judge and subsequently the Board denied most of these claims, finding that Jones had been reinstated, had not lost future wages, and that his emotional distress predated the removal. Only $10,000 in compensatory damages for emotional distress was awarded.

Jones, representing himself without a lawyer, then petitioned the Federal Circuit for review of the Board’s decisions on his compensatory and consequential damages claims. In his briefs and in a prior motion for sanctions, he included case citations that the government’s response brief flagged as apparently non-existent — the result, the court found, of the use of artificial intelligence to generate legal research.

The AI Issue

The central AI-related legal question was whether a pro se litigant who submits briefs containing fabricated AI-generated case citations to a federal appellate court is subject to sanctions — and whether the court will treat such conduct as sanctionable regardless of the litigant’s professional status or claimed ignorance of the error. The court also had to decide what remedy to impose on a first occurrence, and what warning to issue for future filings.

What the Court Decided

• The court identified multiple citations in Jones’s briefs and motion for sanctions that did not appear to exist, including Riley v. Dep’t of Veterans Affs., 64 M.S.P.R. 42 (1994) and Lipscomb v. Dep’t of the Army, 301 F.3d 1356 (Fed. Cir. 2002) — both apparently fabricated by AI.
• The government’s response brief identified a list of additional apparently fraudulent citations that Jones did not attempt to explain or justify in his reply.
• The court held that the use of false, AI-generated citations is sanctionable conduct, stating this expressly on the face of the opinion.
• The court warned Jones that repetition of this conduct can result in serious consequences, potentially including dismissal of his petition for review.
• On the merits, the court affirmed the denial of most damages, but vacated and remanded on the narrow issue of lost future wages and backpay claims to be consolidated with the enforcement proceedings.

“The use of these false citations is sanctionable, and repetition of this conduct can result in serious consequences, potentially including dismissal of his petition for review.”

— U.S. Court of Appeals for the Federal Circuit, Per Curiam, May 13, 2026

The India Angle

Indian Law Equivalent

Indian courts rely on the doctrine of contempt of court and the provisions of the Contempt of Courts Act, 1971, to address misleading or false representations to the court. Additionally, Order VI Rule 16 of the Code of Civil Procedure, 1908 empowers courts to strike out pleadings that are scandalous, frivolous, or vexatious. The Supreme Court has repeatedly held that advocates and parties have a duty of candor to the court — submitting non-existent case citations would squarely violate this duty and could attract contempt proceedings or striking of pleadings.

Bar Council Rules

Under Rule 15 of the Bar Council of India Rules (Standards of Professional Conduct and Etiquette), an advocate must not mislead the court by citing incorrect case law or suppressing material precedents. Submitting AI-generated fictitious citations would constitute professional misconduct under this rule, exposing advocates to disciplinary action before the relevant State Bar Council and the Bar Council of India, including suspension or removal from the rolls under the Advocates Act, 1961.

Practical Advice for Indian Advocates

• Never file a brief using AI-generated citations without manually verifying each case in authenticated databases such as SCC Online, Manupatra, or the Supreme Court’s official website — hallucinated citations are invisible errors that will only surface when the court or opposing counsel looks them up.
• Even pro se litigants in India are held to the standard of not misleading the court; if you assist or coach a self-represented party in preparing submissions, verify their citations before filing.
• Maintain a simple pre-filing checklist: paste every citation into a verified legal database, confirm the page number and holding, and print the case header before filing — this five-minute step can prevent contempt proceedings or disciplinary action.

Quick Takeaways

• The Federal Circuit confirmed that AI-fabricated citations are sanctionable whether submitted by attorneys or pro se litigants — professional status provides no shield.
• The court chose a graduated response on first occurrence (warning + express threat of dismissal) rather than immediate sanctions, but the language leaves no doubt about the severity of the next violation.
• For Indian advocates, the duty of candor under the Bar Council of India Rules means that submitting unverified AI citations could constitute professional misconduct independent of any contempt action by the court.

---

Deep Dive: Why Pro Se Litigants Are Not Exempt from AI Citation Rules

One of the persistent misconceptions in discussions about AI hallucination liability is that the problem belongs exclusively to lawyers. The assumption is that courts will be lenient with self-represented parties who lack legal training, who may not know how to evaluate the reliability of an AI’s output, and who are navigating a complex system without guidance. Jones v. Department of the Army firmly rejects that assumption, at least at the Federal Circuit level.

The court’s per curiam opinion identifies the fabricated citations in plain language — naming the non-existent cases and noting that they “do not appear to exist.” The phrase “apparently resulting from the use of artificial intelligence” appears in the opinion without any note of sympathy or mitigation. The court goes further: it observes that Jones “does not attempt to explain or justify the inclusion of these citations” in his reply brief, treating the silence as an aggravating factor rather than a neutral one. The warning that follows — that repetition “can result in serious consequences, potentially including dismissal” — is as stark as any warning issued against a represented party in the AI citation cases that have proliferated since Mata v. Avianca in 2023.

This matters for several reasons. First, it signals that the Federal Circuit views fabricated citations as a systemic problem requiring consistent enforcement regardless of representation status. Courts have enough difficulty managing their dockets without having to cross-check every citation in every brief; once a court discovers that a party is submitting non-existent cases, it cannot verify anything else in that party’s filings with confidence. The institutional harm is the same whether the filer is a licensed attorney or a whistleblower representing himself. Second, the ruling places a clear verification burden on anyone who uses AI for legal research — including individuals who turn to chatbots and AI legal assistants because they cannot afford a lawyer.

Third, and most practically, the decision illustrates that the consequences of AI hallucinations in federal litigation are no longer merely a professional ethics problem — they are a litigation strategy risk. Jones’s underlying claims on backpay and future wages actually had merit: the court vacated and remanded those issues precisely because the enforcement proceedings contained contradictory factual findings. Had the AI citation issue metastasized into a dismissal, Jones would have lost any ability to vindicate those legitimate claims. The irony is that AI may have been used to strengthen a case that had a real chance of partial success, and in doing so, it created a vulnerability that could have destroyed that chance entirely.

For Indian practitioners advising clients who want to use AI to manage litigation costs, this case is a crucial reference point. The duty not to mislead the court is not a professional monopoly held by lawyers — it extends to everyone who appears before the court, however they are situated. The verification obligation that comes with AI-assisted legal research is not a formality; it is the minimum required to meet that duty.