Appeals Court Rules Law Firms Responsible for AI-Generated Errors in Legal Filings

A federal appeals court ruled on November 3 that attorneys bear full professional responsibility for the accuracy of AI-generated content submitted in legal filings — rejecting the argument that reliance on an AI tool's output constitutes a reasonable excuse for factual errors or hallucinated citations. The decision extends a now-consistent body of case law confirming that AI is a tool, not a co-counsel, and that professional responsibility rules do not create an AI exception.

The ruling is significant not because it is surprising — several courts have reached the same conclusion — but because it comes from an appeals court, which means it carries precedential weight across the jurisdiction and closes off lower-court experimentation with more lenient standards.

What the Court Ruled

According to JD Journal's reporting, the appeals court upheld sanctions against attorneys who submitted filings containing AI-generated factual claims and case citations that turned out to be fabricated. The attorneys argued that their reliance on the AI output was reasonable — that they had no way to know the model had hallucinated — and that sanctions should therefore be reduced or vacated.

The court rejected that argument on two grounds. First, attorneys have an independent duty to verify the accuracy of everything submitted over their signature; that duty predates AI tools and applies regardless of where the content originated. Second, the existence of widely documented hallucination problems in AI legal drafting tools — extensively covered in legal trade press and multiple prior court sanctions orders — means attorneys cannot claim ignorance of the risk. At this point, using an AI legal tool without verification is not a mistake; it is a choice.

The court awarded attorney fees to opposing counsel, with the sanctioned attorneys and their firm jointly and severally liable.

The Growing Sanctions Record

This ruling extends a pattern that began with the Mata v. Avianca case in 2023, when ChatGPT-generated fake citations ended up in a Southern District of New York filing. Since then, federal and state courts have issued sanctions in multiple jurisdictions, bar associations have issued guidance, and the Judicial Conference of the United States has considered AI disclosure requirements.

What has emerged over three years is not a single rule but a consistent principle applied across different court systems: AI-generated content is attorney work product the moment it is submitted. The attorney who submits it is responsible for it. The AI vendor has no professional obligations; the attorney has all of them.

This November ruling is notable for two reasons: it comes at the appeals level, giving it broader precedential force, and it involves attorney fees — a sanction that falls on the law firm's finances, not just its reputation. Attorney fee awards are often the mechanism that changes firm behavior where admonishment does not.

Impact on Legal Practice

Law firms using AI drafting tools — which is now most of them — need to audit their verification workflows. The question is not whether to use AI assistance; it is whether the firm has a documented, consistent process for a human attorney to verify every factual claim and citation before it goes into a filing.

Some firms have invested in AI-specific review workflows: a second attorney spot-checks AI-generated content against primary sources before sign-off. Others are using verification-layer tools — software that cross-references citations against legal databases. The appeals court ruling makes clear that neither the tool nor the workflow is the court's concern; the attorney's signature is.

Malpractice insurers will read this ruling carefully. Several carriers have already begun asking about AI use in legal drafting during policy renewals. A ruling that explicitly holds firms jointly and severally liable for AI-generated filing errors strengthens the argument for mandatory AI-use disclosures in malpractice coverage applications.

Clients who pay attorney fees as a sanction may also have recourse against their counsel. If a firm's AI workflow failure resulted in an adverse ruling or required remedial work, the client's claim that the firm breached its duty of competence becomes easier to make.

What to Watch

Watch whether the Judicial Conference moves forward with mandatory AI disclosure requirements in federal filings — a proposal under consideration for the past year. If disclosure becomes required, courts gain visibility into which filings used AI assistance, creating a data layer that will likely produce more sanctions proceedings, not fewer. Also watch whether state bar associations begin issuing ethics opinions that go beyond guidance into explicit disciplinary standards for AI verification failure.

The practice of law is built on the principle that an attorney's signature means something. Courts are making clear, one ruling at a time, that AI does not change what that signature means.