In the U.S. District Court for the Northern District of Georgia, defense counsel representing Bank of America in *Ringer v. Bank of America, N.A.* found themselves in an uncomfortable spotlight last November. A federal judge, spotting what the court described as "multiple inaccurate quotations or citations" in a motion to dismiss, issued an Order to Show Cause — the judicial equivalent of a raised eyebrow backed by the threat of sanctions. According to a defendant's response filed November 25, 2025, the firm's answer was unambiguous: no generative AI tool touched this brief.
The sworn declaration, filed under penalty of perjury, describes a more mundane and arguably more embarrassing origin for the errors. An attorney assembled a working document that blended copied excerpts, paraphrased summaries, and personal research notes — a patchwork that was then folded into the final filing without a proper citation check. Some paraphrases ended up dressed as direct quotations. Some case citations were simply wrong. The attorney took full responsibility and outlined remedial steps: stricter verification protocols, secondary review requirements, and additional attorney training. As of the filing date, the court had not yet ruled on whether sanctions would be imposed.
The firm's denial matters because it cuts against a narrative that has been building in courtrooms nationwide. Judges have grown so accustomed to seeing AI-hallucinated citations — fabricated cases that sound plausible but do not exist — that citation errors now trigger immediate suspicion of AI misuse. That suspicion is not unfounded. As reported by the *National Law Review* and confirmed by court records, two Colorado attorneys were sanctioned under Federal Rule of Civil Procedure 11 after submitting briefs containing nearly 30 defective AI-generated citations. In Wyoming, three attorneys from a major plaintiffs' firm — Morgan & Morgan — were fined after their motion included eight fabricated case citations produced by AI tools; the drafting attorney was hit with a $3,000 fine and supervising counsel faced additional sanctions for inadequate oversight. A California state appellate court went further, fining one attorney $10,000 for filing an appeal stuffed with dozens of fake ChatGPT-generated quotations.
What ties these cases together is not the technology itself but the professional obligation that predates it. Model Rule 3.3 requires candor toward the tribunal; Model Rule 1.1 requires competence. Neither rule has an AI exemption, and judges have been explicit on that point: using a generative AI tool does not transfer the attorney's verification duty to the machine. Courts have required attorneys to refund fees, cover opposing counsel's costs, and in some instances submit to disciplinary review — consequences that fall on the lawyer, not the software. The Georgia case, whether or not it results in sanctions, underscores that the same accountability applies when the culprit is old-fashioned human carelessness rather than algorithmic hallucination.
The broader takeaway for the legal profession is practical and uncomfortable. Firms rushing to adopt AI drafting tools face obvious risks if they skip verification. But the Georgia episode demonstrates that firms *not* using AI face a different problem: courts now treat citation errors as presumptively suspicious, meaning attorneys must affirmatively prove their mistakes were human. That is a new and uninvited burden. Across the profession, firms are updating internal policies — citation verification checklists, mandatory secondary review, explicit AI-use guidelines — less out of ethical awakening than out of self-preservation. The judicial system is watching, and it is not in a forgiving mood.
