When AI Hallucinates the Law: Sedano and Nolan Show the Cost of Generated Legal Writing

March 17, 2026

When AI Hallucinates the Law: Sedano and Noland Show the Cost of Generated Legal Writing

What clients need to know about AI-generated citations, sanctions exposure, and the safeguards LFLM uses to protect your files.

Generative AI can speed up routine legal work, such as drafting e-mails and summarizing documents.  However, when it invents quotations or cites cases that do not exist, courts are imposing real sanctions and, in some instances, referring counsel to the State Bar.  Two 2025 decisions – Sedano v. Live Action General Engineering Inc. and Noland v. Land of the Free, L.P. – illustrate how quickly unverified AI-assisted writing can turn into a credibility and cost problem.  The technology may be new, but the professional duties of accuracy and candor are not.

The issue is straightforward; a filing that relies on fabricated or mismatched authorities can waste judicial resources, misstate the law, and undermine the integrity of the process.  Courts and the WCAB already have tools to address this conduct through sanctions and procedural enforcement. [1]

Sedano: WCAB Sanctions for a Defective Petition

In Sedano v. Live Action General Engineering Inc., the WCAB reviewed a Petition for Reconsideration that, in the panel’s words, appeared not to have been meaningfully reviewed before filing.  The petition lacked proper record citations, raised issues never tried, and most importantly, cited authorities that were irrelevant, non-citable, or non-existent.

On June 30, 2025, the Board granted reconsideration for the limited purpose of issuing a Notice of Intention to impose sanctions under Labor Code section 5813 and 8 C.C.R. section 10421, inviting defendant to address whether AI was used and why the defects persisted even after the WCJ’s report.  On July 29, 2025, the WCAB affirmed the Findings and Award and imposed joint and several sanctions on the employer, insurer, TPA, and defense counsel.

The lesson is not limited to one petition.  A defective petition can create sanctions exposure even where the underlying merits fail.  More importantly, it can distribute the consequences across multiple defense stakeholders when the filing appears to reflect a breakdown in verification and supervision.

Noland: Fabricated Citations at the Appellate Court

In Noland v. Land of the Free, L.P., the Court of Appeal affirmed summary judgment, but was published primarily to address problems in counsel’s briefing.  The court found that many quoted passages did not exist in the cited cases, several citations did not support the propositions offered, and at least one case could not be located at all.  The court relied on California Rules of Court, rule 8.276, and emphasized that the attorney has a non-delegable duty to verify authorities.  It imposed sanctions, directed that the opinion be served on the client, and ordered notice to the State Bar.

From Noland, we see that courts at all levels are now actively checking quotations and citations.  When authorities are fabricated or do not match the proposition, the response can include sanctions and reputational damage that extends far beyond a single case.  Those outcomes can complicate future credibility in negotiations and litigation, particularly in high-volume settings where consistency is key.

Mitigating Risk of AI Hallucination

AI hallucinations are not simply an ethics issue.  They are a litigation risk with predictable costs.  A defective petition or brief can forfeit issues, trigger sanctions, and require expensive re-work.  Even when the dollar figure of sanctions is modest, the soft costs are real, including delay, distraction, and diminished credibility.

The safeguards against AI hallucination are simple.  When human judgment is required, all AI outputs must be treated as untrustworthy until verified.  Any document that contains a citation must be personally checked by an attorney for accuracy.  In petitions or briefs especially, every quoted passage must be checked to ensure it appears in the cited source exactly as represented, and each authority must be reviewed to confirm it actually supports the proposition for which it is cited.  Finally, all practitioners should have a working knowledge of the Labor Code, California Code of Regulations, and California Rules of Court to ensure filings meet candor and accuracy standards.

The Bottom Line

While AI has demonstrated value in accelerating routine tasks, it cannot replace verification by a trained and licensed legal professional.  The consistent message from Sedano and Noland is that accuracy is not optional, and courts will impose consequences when filings contain invented quotations, unsupported citations, and fabricated cases.  For employers, carriers, and administrators, as AI tools continue to be integrated in all levels of workers’ compensation practice, the safest approach is to insist on verification practices that protect the integrity of the case and your position before the bench.