A federal judge in Manhattan is warning that law firms' growing reliance on artificial intelligence could put clients at risk and leave junior lawyers without the training they need to do the job. U.S. District Judge Arun Subramanian raised the alarm after citation errors surfaced in a copyright infringement case, even as he decided not to punish the attorneys involved.
Subramanian, who sits in Manhattan, questioned whether unchecked AI use was behind the citation mistakes in the case, according to Reuters. He ultimately declined to sanction the attorneys over the errors, but used the episode to issue a broader caution to the legal profession: AI-generated legal work must be carefully checked for accuracy before it ever reaches a judge's desk.
The errors traced back to a filing from Shapiro Arato Bach, which had used AI tools built by Harvey, a technology company, to help prepare the document. Cynthia Arato, an attorney for defendant Foundation Media, apologized for the mistakes in the filing. She said the problems stemmed from another lawyer who had prepared the initial draft and finalized it before it was submitted to the court.
Firm Says Error Was Isolated, Not Intentional
In a declaration cited by Reuters, Arato characterized the failure to check the AI-generated citations as an isolated error made without bad faith or any intent to mislead or deceive the court. Shapiro Arato Bach said the unnamed lawyer responsible acted contrary to the firm's own policies and training, and the firm said it is now implementing further safeguards against AI misuse. Harvey and Foundation Media did not respond to requests for comment, per the same report.
Subramanian said the episode should serve as a wakeup call for lawyers across the profession, not just the firm involved in this case. The judge's broader warning focused on the risks AI can pose to clients and on how reliance on the technology may affect junior lawyers' training.
A Shared Responsibility as Technology Reliance Grows
Subramanian warned that reliance on AI could leave junior lawyers without the training they need.
The warning lands as AI adoption climbs across the legal industry but formal guardrails lag behind. A survey of more than 2,800 legal professionals found personal generative AI use at work rose to 31% in 2025, up from 27% the year before, according to the American Bar Association. The same survey found adoption climbing to 39% at firms with 51 or more lawyers, compared with roughly 20% at smaller shops, with 54% of legal professionals using AI to draft correspondence.
Policies and Training Haven't Kept Pace
Despite that growth, oversight remains thin. A 2026 survey found that 43% of respondents said their firm had no formal AI policy and no plans to create one, while only 9% had a written policy that was actually enforced, the North Carolina Bar Association reports. The same survey found that 54% of respondents said their firms provided no training on responsible AI use and had no plans to start, even though Clio's 2026 survey found 71% of solo practitioners and 75% of small firms had adopted AI in some form.
Hesitancy persists for good reason. Three-quarters of lawyers surveyed by the ABA Journal said concerns about AI-generated hallucinations were the main thing holding them back from wider implementation, and the American Bar Association's report attributed uneven adoption to a mix of practical, cultural and economic factors, including restrictive internal policies and pressure to prove return on investment. The ABA Journal has reported that firms using AI successfully tend to pair it with human oversight and guardrails rather than letting the technology run unchecked.
The legal exposure for getting it wrong is real. The U.S. District Court for the Northern District of California has stated plainly that anyone who chooses to use an AI tool remains responsible for everything filed with the court under Federal Rule of Civil Procedure 11. Separately, the U.S. District Court for the Eastern District of Texas has issued a standing order requiring litigants to review and verify any computer-generated content to ensure it complies with Rule 11, Local Rule AT-3 and other standards of practice. Legal analysis published by the American Bar Association Business Law Today notes that violations of Rule 8 can lead to dismissal of a complaint, while violations of Rules 11 and 26 can trigger a range of sanctions against the attorneys involved.
