AI citation sanctions need to hurt more
California’s latest $10,000 fine shows the fake-citation problem is no longer a one-off embarrassment; it is a repeated professional failure that courts should punish publicly and hard.
An AI-assisted editorial, reviewed by a human before publishing. It reasons over our own tracker data (and cited context). A point of view, not legal advice.
The usual line is that AI citation blunders are embarrassing but fixable. That is too soft. A California appeals court just fined a lawyer $10,000 after he filed ChatGPT-generated fake case quotes and citations, and that belongs in the same category as any other filing that tells the court a lie.
AI does not get a softer rule because the lie was generated faster.
This is not a novelty offense
The California fine is not standing alone. Your own database already tracks 724 AI-related sanctions, with California at 57 and New York at 63. That is what a pattern looks like, not a one-off accident.
The recurring facts are plain. In Nevada, Jan Tomasik and Daniel Mann filed a brief with bogus citations generated by ChatGPT; Judge Hardy ordered them removed, referred them to the Nevada bar, and fined them. In another case, Walters filed papers with hallucinated citations and was sanctioned. In a U.S. sanctions matter from 2023, two lawyers and their firm were hit with $5,000 fines each after a brief with fake cases. In the Southern District of New York, a lawyer was challenged over seven apparently fictitious cases, asked ChatGPT to summarize them, and the court fined the two lawyers and their firm $5,000. In the Illinois case reported by Law.com, the appeals court went further still, imposing a $15,000 fine and saying it hoped to deter the misuse of AI in case briefs.
Courts are already saying the quiet part out loud
This is the part lawyers should stop pretending not to hear. Courts are no longer treating these filings as a charming sign of confusion in a new tool. The Illinois court said the misuse was unacceptable and criticized a sloppy filing that misstates the law and creates more work for the court. Reuters reported that Connecticut’s highest state court warned lawyers they remain responsible for the accuracy of filings produced with generative AI. That is the right frame: responsibility does not migrate just because the drafting tool got faster.
There is also a broader market reality that should embarrass anyone still calling this experimental. Legal AI tools are everywhere in the vendor catalog, and the tools getting the most attention right now include Claude for Legal, CoCounsel, and ContextGem. At the same time, one report from Law.com says firms are pouring money into AI while hiring fewer entry-level lawyers. If senior lawyers are going to let software do the first pass, they had better be prepared to do the second pass themselves.
The counterargument is real, but it does not rescue bad filing practice
The strongest defense is that lawyers are under pressure and AI can make research and drafting faster. Fine. Harvey’s own marketing says its tools support research, contract review, litigation work, and end-to-end agents. That may help a lawyer move faster. It does not excuse filing a brief with fake authorities.
Speed is not a defense when the document goes to court. The problem is not that a lawyer used AI. The problem is that the lawyer filed AI output without verification. That is not a complicated ethical question. It is basic professional negligence dressed up in new software.
What courts should do now
The answer is not more hand-wringing and not some special AI exception. Courts already have the tools, and the recent sanctions show they know how to use them. The next step is to make the penalty predictable: real money, public orders, referrals where appropriate, and no pretending that a fake citation is a clerical hiccup.
A lawyer who hands a judge a brief containing hallucinated authorities should expect a sanction that stings, a record that follows them, and a message the bar cannot miss. The profession learned this lesson long ago with other forms of false filing. AI does not get a softer rule because the lie was generated faster.