We have all seen the headlines about lawyers who are being sanctioned for citing cases that do not exist. For a while, those stories felt like cautionary tales of isolated mistakes that resulted in modest penalties. However, a recent decision out of Oregon might be a precursor to more serious sanctions and penalties for lawyers who submit hallucinated pleadings.

In this case, the court confronted something far more serious than a stray hallucinated citation. The plaintiffs’ filings contained fifteen non-existent cases and multiple fabricated quotations attributed to real authorities. And when opposing counsel flagged the issue, the response was not an immediate correction or expression of remorse. Instead, the court found a pattern of evasion, half-measures, and continued reliance on unsupported legal assertions.

The result was severe. The court imposed attorneys’ fees and monetary sanctions totaling approximately $110,000 and, most notably, dismissed the plaintiffs’ claims with prejudice. That last piece is what should get every lawyer’s attention. This was not just a sanctions order. It was a case-ending event.

The court framed the issue in familiar terms. Rule 11 requires a “reasonable inquiry” into both facts and law before filing anything with the court. Generative AI does not change that obligation. If anything, it heightens it. As the court emphasized, large language models are not designed to tell the truth. They are designed to produce plausible language. When lawyers outsource legal research to a tool that predicts words rather than verifies authority, the risk is not just error. It is fiction presented as law. And courts are losing patience with it.

What makes this case especially striking is not just the volume of hallucinations, but the failure to fix the problem. The court made clear that simply deleting fake citations after being caught is not “corrective action.” The duty is to verify, before filing, that the law you cite actually exists.

There is a broader point here for practicing lawyers. We tend to think of AI-related sanctions as technology problems. They are not. They are competence problems. The same rules that prohibit citing bad law prohibit citing imaginary law. The tool does not matter. What has changed is the scale of the risk.

If there is a takeaway from this decision, it is a simple one. You cannot delegate judgment to a junior lawyer, a research assistant, or a machine that does not know the difference between actual authority and fiction. Because when the citations are fake, the consequences are very real.