On Thursday, Judge Jennifer Choe-Groves of the Court of International Trade issued an opinion in the matter bearing the caption Disruptive Resources, LLC v. Ballistic Barrier Prods. Inc. (D. Del.).
Before the Court lies a patent-infringement dispute between Plaintiff Disruptive Resources and defendants Ballistic Barrier Products Inc. and Mayday Security Solutions LLC. Following disclosure by Plaintiff’s counsel that the Joint Claim Construction Brief previously submitted contained several mistakes arising from the use of generative artificial intelligence (“AI”), the Court directed the attorney to provide a written showing of cause explaining why sanctions under Federal Rule of Civil Procedure 11 should not be imposed for the manner in which counsel conducted the filing.
In the letter submitted to the Court, counsel recounted the sequence that led to the disclosure of AI-driven inaccuracies. He noted that portions of his reply brief in support of Plaintiff’s proposed claim constructions included an erroneously proposed interpretation that diverged from the constructions Plaintiff had advocated during the claim construction proceedings.
Counsel admitted employing the AI system “Strongsuit” to draft an initial outline of Plaintiff’s reply brief and to extract deposition excerpts that favored Plaintiff’s position on claim construction. He used targeted prompts to complete an initial draft of the reply brief that incorporated extrinsic evidence referenced in the deposition.
During the editing process, counsel moved between sections rapidly and mistakenly believed he had completed edits across the entire brief, when in fact a single one-page section proposing the incorrect construction had been overlooked. He further acknowledged that he did not perform a final, comprehensive review of the entire brief and that some arguments were not fully assessed.
Upon recognizing the erroneous statements, counsel notified opposing counsel and proposed amending the Joint Brief to add footnotes addressing the mistaken constructions. The parties subsequently filed a Joint Letter informing the Court of these issues. Counsel contends that his actions did not violate Rule 11 or professional conduct rules because the reply brief was not advanced for an improper purpose, the claims and legal contentions rested on authentic, existing authority, and all factual contentions were supported by evidence. He characterizes his missteps as efforts at advocating positions that the Court ultimately did not adopt.
The Court notes that the Joint Letter shows that two portions of the Joint Brief carried a concluding statement proposing an incorrect claim construction, and a third portion was withdrawn entirely because it contained more pervasive errors. Plaintiff sought to strike the problematic portion of its reply brief, prompting Defendants to strike their subsequent briefing that responded to the AI-generated, erroneous arguments raised by Plaintiff. This sequence demonstrates that the flawed constructions not only affected the precision and persuasiveness of Plaintiff’s legal arguments but also forced opposing counsel to spend resources addressing arguments that were ultimately excised from the record.
The deployment of generative AI within the legal profession is expanding at a rapid pace, a trend evidenced by counsel’s admission that he used large language models to alleviate workload burdens for his smaller firm. Nevertheless, the practice of law remains governed by strict ethical standards, and the use of AI does not dispense with these obligations. In the Court’s view, counsel’s letter and the remarks at the hearing reflect remorse for the consequences of using AI without diligent review and a recognition that a higher standard of careful advocacy is required than the one that produced the erroneous filings.
These facts stand in contrast to other instances where AI-generated material has produced fictitious citations or invented judicial opinions. See Mata v. Avianca, Inc. (S.D.N.Y. 2023) (where an attorney submitted non-existent opinions with invented quotes and citations created by an AI tool, and persisted in relying on the fake authorities despite orders questioning their existence). In this matter, counsel accepted responsibility for his mistakes, did not evade accountability, and prioritized candor before the Court and opposing counsel by promptly disclosing the errors. By contrast, Mata involved conduct where the attorney knew, prior to submitting the materials, that the fake text could not be verified but remained silent.
The Court expects that the events and this Opinion will serve as a cautionary lesson and a training moment for counsel going forward. It cautions that any future AI-related missteps in this case may lead to sanctions.
After weighing the circumstances presented by counsel and recognizing the gravity of submitting AI-drafted legal arguments without proper review, the Court has decided not to sanction counsel for his conduct in this instance….
