Hardy v. GCCARD: Federal Court Strikes AI-Hallucinated Motion After Prior Warning, Calls Out Repeat Offender | Advocate Prakhar

Case at a Glance

Background

Gregory Hardy is a pro se plaintiff who has filed multiple civil rights lawsuits in the Eastern District of Michigan. In Hardy v. Whitaker (related case, 24-11270), the court issued a report and recommendation in March 2026 recommending that Hardy be sanctioned, in part because he had used fake citations in a motion for summary judgment. The R&R quoted the Sixth Circuit and noted that “Courts and opposing counsel are burdened by the careless use of AI by pro se filers.” Hardy apologised for his “inclusion of inaccurate citations generated by an AI tool” in his objections to that R&R.

Within one month, Hardy filed an identical motion to consolidate in three of his related cases — including the present GCCARD case. The consolidation motion cited Berndt v. Tennessee for the proposition that a court must address a litigant’s mental capacity before dismissing for failure to follow rules. The Berndt opinion contains no such language and does not address mental-capacity-related dismissals. The motion also cited United States v. Frazier for the proposition that a judge may not act as an “amateur psychologist” — a phrase that appears nowhere in that criminal case.

The AI Issue

Hardy’s conduct represents what courts are increasingly calling a pattern of AI misuse: apologise for one instance, repeat it in the next filing. The Eastern District of Michigan court was able to compare the motion to consolidate with Hardy’s prior apology in the related case — the same AI error, the same type of fabricated quotation, just one month later. The Sixth Circuit had also just held (in Whiting v. City of Athens Tennessee, March 13, 2026) that even a single fake case can be sanctionable and that “citing fake cases can harm the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions.” The court cited this authority as the escalating legal backdrop for Hardy’s repeat conduct.

What the Court Decided

• Motion to consolidate struck for containing fabricated quotations from Berndt v. Tennessee and United States v. Frazier.
• Motion characterised as frivolous because it falsely claims the court has a duty to address Hardy’s mental capacity.
• Renewed motion to compel discovery denied for failure to address applicable legal standards (proportionality under Rule 26(b)) and for citing additional fabricated authorities including “Williams v. Reed, 605 U.S. __ (2025)” for a proposition (“17 Quarters of Silence” as a per se constitutional violation) that does not appear in the opinion.
• Court cited Sixth Circuit’s March 2026 ruling that even a single fake case can be sanctionable under Rule 11 and inherent authority.
• Court noted Hardy had rejected help from the Federal Pro Se Legal Assistance Clinic when it gave him advice he did not like.

“[E]ven a single fake case can be sanctionable” and “citing fake cases can harm the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions.”

— Sixth Circuit Court of Appeals, Whiting v. City of Athens Tennessee, 2026 WL 710568 (Mar. 13, 2026), cited in Hardy v. GCCARD

The India Angle

Indian Law Equivalent

In India, repeat fabrication of citations across multiple proceedings would be treated under the Contempt of Courts Act, 1971. Section 2(c) defines criminal contempt as conduct that “scandalises or tends to scandalise, or lowers or tends to lower the authority of, any court.” Filing fabricated citations that put false words in judges’ mouths — attributing to real cases holdings they do not contain — scandalises the court whose name is invoked as author of the fabricated holding. The Supreme Court has broad powers under Article 142 to ensure that litigants who abuse the process are appropriately dealt with, including through adverse cost orders, dismissal, and referral to the Bar Council.

Bar Council Rules

The BCI’s disciplinary framework under Section 35 of the Advocates Act, 1961, treats repeat misconduct more seriously than isolated offences. An advocate who apologises for AI-hallucinated citations in one case and then files fresh hallucinated citations in the next case within weeks demonstrates that the apology was performative rather than genuine. The Bar Council’s authority to impose penalties ranging from censure to debarment would be exercised with increased severity in such circumstances.

Practical Advice for Indian Advocates

• An apology for AI misuse in one case does not provide immunity for AI misuse in the next — courts and disciplinary bodies track patterns across cases, and repeat conduct escalates the consequences dramatically.
• The Sixth Circuit’s holding that even a single fake case is sanctionable is a powerful signal: the threshold for sanction is not high volume or systematic deception — any single AI-hallucinated citation can trigger professional consequences.
• If a court identifies AI-hallucination errors in your filing, implement a verification protocol immediately — not just for the affected case, but for all pending and future filings. A structural failure that produced one error will likely produce more unless the process is changed.

Quick Takeaways

• Prior apologies for AI hallucinations do not reduce the courts’ authority to sanction fresh instances — courts will hold apologies to account when the same conduct recurs.
• The Sixth Circuit has confirmed that even a single fake case is sanctionable under Rule 11 and inherent authority — there is no de minimis threshold for AI hallucination misconduct.
• Pro se litigants who reject available legal assistance (such as pro se clinics) and then file AI-hallucinated briefs receive diminished sympathy from courts that tried to help them.

Deep Dive: The Repeat Offender Problem — Why Apologies Without Process Change Are Not Enough

The Hardy cases — there are at least three related ones — present the clearest example in the reported case law of what happens when a litigant treats an AI-hallucination apology as a legal defence rather than as a commitment to change. Hardy apologised in Hardy v. Whitaker and was given the benefit of that apology. One month later, he filed identical fake quotations in Hardy v. GCCARD and Hardy v. Jones. The courts in all three cases connected the dots.

The explanation for this pattern is probably not bad faith in the traditional sense. Hardy is a pro se litigant who apparently uses an AI tool to generate legal arguments, and has not changed his workflow after his apology — he is still feeding case facts to the AI and submitting its output without reading the cited cases. The apology was genuine in the moment, but the underlying process that produces the errors remained unchanged.

This has implications for how courts and disciplinary bodies should approach AI-hallucination sanctions. Monetary sanctions alone may not change the behaviour of a pro se litigant who lacks the resources to pay them and who continues to generate filings using the same AI tool. More effective sanctions are those that address the process: mandatory legal clinic consultations, requirements to submit verified case citations with digital copies, or restrictions on filing without prior court approval. These process-based sanctions are more burdensome than monetary ones but more likely to actually prevent the error from recurring.

For advocates in India who supervise junior lawyers or manage pro se client matters, the Hardy cases demonstrate that AI verification failures are often systemic rather than isolated. If a junior’s filing contains one AI-hallucinated citation, check every other recent filing that junior has submitted. The same workflow failure that produced one error has almost certainly produced others. A comprehensive audit, followed by a firm-level process change, is the appropriate response — not an isolated correction and an apology.