The AI Double Standard: When Federal Judges Play by Different Rules Than the Attorneys Before Them
Introduction
Here’s a story that would be comedic if it weren’t so troubling: In July 2025, New York attorney Steven Feldman found himself in hot water with Judge Katherine Polk Failla for using AI to generate a brief filled with fake citations. His response to her order to show cause? Another AI-generated document that misattributed quotes and contained flowery prose about ancient scribes and Ray Bradbury’s Fahrenheit 451. Meanwhile, just days later in New Jersey, Judge Julien Xavier Neals quietly withdrew his own opinion after defense counsel pointed out it contained multiple nonexistent quotes and misstated case outcomes—problems strikingly similar to those plaguing attorney AI use, though the judge’s errors have not been officially attributed to AI.
The irony is as thick as the judicial robes themselves. While attorneys face sanctions, fines, and professional embarrassment for AI mistakes, federal judges who make identical errors simply issue quiet retractions and move on. This tale of two standards reveals a fundamental problem in our justice system’s approach to artificial intelligence.
Background: The Attorney AI Crackdown Intensifies
The legal profession’s AI reckoning began with the infamous Mata v. Avianca case in 2023, where attorneys submitted completely fabricated case citations generated by ChatGPT. Since then, courts have become increasingly vigilant about attorney AI use, implementing strict disclosure requirements and harsh sanctions for violations.
The Feldman case in Judge Failla’s courtroom represents this trend’s latest chapter. After filing a brief containing nonexistent citations in a trademark infringement case¹, Feldman faced an order to show cause. His response acknowledged that errors “resulted from sophisticated AI hallucination mechanisms rather than intentional misconduct,” but Judge Failla wasn’t buying it. She noted the response’s dramatic shift in writing style compared to Feldman’s subsequent letter, which was riddled with typos—suggesting the polished response itself was AI-generated².
“Mr. Feldman must know how to verify that a case exists on Westlaw without the added benefit of AI tools,” Judge Failla wrote, embodying the judicial expectation that attorneys should know better than to rely on unverified AI output³.
Key Trends: The Emerging Judicial Double Standard
The Silent Retraction vs. Public Sanctioning The contrast couldn’t be starker. While Judge Failla publicly dressed down Feldman with Shakespeare quotes and scheduled a conference to hear from him directly, Judge Neals in New Jersey simply withdrew his problematic opinion with a brief text order stating it was “entered in error”⁴. No show cause orders, no conferences, no explanations to the parties about how an entire judicial opinion came to contain fabricated quotes and mischaracterized case law.
Identical Errors, Different Consequences The errors in both cases are remarkably similar, though their causes may differ. Feldman’s brief contained fake case citations and misattributed quotes—problems he acknowledged resulted from AI use. Judge Neals’ opinion in In re CorMedix Securities Litigation contained strikingly similar issues: nonexistent quotes attributed to real cases, misstated case outcomes (claiming motions to dismiss were denied when they were actually granted), and even attributed statements to defendants that they never made⁵. While the source of Judge Neals’ errors remains unspecified, the pattern mirrors common AI hallucination problems. Yet only the attorney faces potential sanctions.
The Accountability Gap Defense attorney Andrew Lichtman’s letter to Judge Neals catalogued a stunning array of errors: quotes that “are not contained in that case,” outcomes that were completely reversed, and fabricated statements attributed to parties⁶. If an attorney had submitted such a brief, sanctions would be inevitable. But when a judge makes identical mistakes, the response is a quiet withdrawal and promise of a corrected version.
The Technology Transparency Divide Courts now routinely require attorneys to certify their AI use and verify all citations. Yet judges operate with complete opacity regarding their own AI usage. There’s no requirement for judicial AI disclosure, no verification protocols, and no accountability when things go wrong.
Practical Tips: Navigating the Minefield
For Attorneys Under Scrutiny Document everything about your AI usage and verification process. Feldman’s case shows that judges expect attorneys to have foolproof verification systems, regardless of access limitations. When he claimed he lacked Westlaw access and relied on co-counsel for verification, Judge Failla was unsympathetic⁷.
Immediate Disclosure Requirements If you discover AI-generated errors in your filings, inform the court immediately. Feldman learned about citation errors on June 22, 2025, but didn’t notify the court until after Judge Failla issued her order to show cause four days later. This delay became another strike against him⁸.
Verification Beyond AI Tools Establish backup verification methods that don’t rely on AI. Judge Failla dismissed Feldman’s explanations about “database migration complications” and “verification gaps,” noting that lawyers should know how to conduct basic legal research without AI assistance⁹.
Professional Responsibility in the AI Age The Feldman case demonstrates that using AI doesn’t absolve attorneys of professional obligations. As Judge Failla noted, choosing to use AI without proper verification “amounts to a series of deliberate choices, not ‘inadvertent’ errors”¹⁰.
Case Examples: The Real-World Impact
The Feldman case showcases how quickly AI problems compound. His original brief contained fake citations, his response to the show cause order contained misattributed quotes, and his subsequent letter was filled with typos—suggesting he alternated between over-relying on AI and abandoning it entirely. Judge Failla scheduled an in-person conference specifically to “hear directly from Mr. Feldman” and give him a chance to prove himself “worthy to carry the stylus once more”¹¹.
Meanwhile, the CorMedix case demonstrates judicial AI errors on a massive scale. Judge Neals’ opinion contained multiple fabricated quotes from real cases, completely reversed case outcomes, and attributed statements to defendants that appear nowhere in the record. Defense counsel’s methodical documentation of these errors reads like a masterclass in legal fact-checking—the kind attorneys are now required to perform but judges apparently are not¹².
The parallel securities case in Outlook Therapeutics adds another layer of irony. Judge Neals’ error-riddled opinion was submitted as “supplemental authority” in that case, potentially spreading the misinformation further before the problems were discovered¹³.
Conclusion
The contrast between Judge Failla’s treatment of attorney Feldman and Judge Neals’ quiet handling of his own errors exposes a fundamental flaw in our justice system’s approach to accountability. Both cases involve similar problems: fake quotes, nonexistent citations, and content that appears unverified. While we know Feldman’s errors stemmed from AI use, the source of Judge Neals’ remarkably similar mistakes remains unaddressed. Yet the attorney faces potential sanctions and public embarrassment while the judge issues a silent retraction.
This double standard undermines the rule of law and creates an atmosphere where judicial authority trumps accountability. If AI transparency and verification are essential for maintaining legal system integrity—and they clearly are—then these requirements must apply equally to everyone in the courtroom, including the person wearing the robes.
The legal profession needs uniform standards that acknowledge AI’s benefits while demanding rigorous verification from all users. Until courts hold themselves to the same standards they impose on attorneys, the AI revolution in law will remain a tale of “do as I say, not as I do.”
Found this analysis of judicial double standards eye-opening? Share it with your legal colleagues and let us know your thoughts on this troubling trend. The conversation about AI accountability in our courts is just beginning, and your voice matters.
#AIinLaw #LegalTech #AIHallucinations #JudicialEthics #GenAI
- Sources:
- Federal court order and Law360 coverage of trademark infringement case, July 2025 (specific case name withheld per publication policy)
- Id. at 2-3
- Id. at 4
- In re CorMedix Inc. Securities Litigation, Case No. 2:21-cv-14020-JXN-CLW, Text Order (D.N.J. July 24, 2025)
Letter from Andrew Lichtman to Judge Julien X. Neals, July 22, 2025, ECF Doc. 123 - Id. at 1-3
- Jack Karp, “NY Judge Wants Atty To Explain Bogus AI Citations,” Law360, July 21, 2025
- Flycatcher Corp., Document 169 at 5
- Id. at 4
- Id. at 3-4
- Id. at 6
- Lichtman Letter at 1-3
- Lauren Berg, “NJ Fed. Judge Pulls Opinion Flagged With Nonexistent Quotes,” Law360, July 24, 2025
