Courts grapple with rise in AI-hallucinated legal filings; over 1,000 instances in past year

Nearly 1,100 legal filings in the U.S. over the past year had false citations, misrepresented legal doctrines and misquoted sources because of errors generated by an artificial intelligence tool, according to an international database.

KXAN spoke on July 23 with Crawford Appleby, a partner at the legal firm Wisner Baum, about these problems.

“This has been happening fairly frequently since Generative AI came on scene years ago,” said Appleby about the AI hallucinations. “But what it means is essentially that attorneys are filing briefs with courts that contain fake legal authority. Usually it’s a fake case … the Generative AI program they were using to help them draft the report documents basically made up or invented some authority that doesn’t exist.”

UT Austin team examining ethics of AI use to release ‘People’s Guide to AI’ in 2026 Appleby is also on Wisner Baum’s AI committee, which the firm says works to develop its policies about AI and “best practices for the responsible use of AI in litigation.”

What does the data show? The firm directed KXAN to a list of cases where a judge, court or tribunal issued a ruling or order that “explicitly found (or implied) that a party relied on [AI-]hallucinated content or material.”

It’s the ongoing project of Damien Charlotin, a senior research fellow at the research university HEC Paris.

According to his last data upload on Aug. 8, there were 1,868 cases worldwide and 1,297 cases in the U.S. When KXAN first learned of the list on July 22, there were 1,784 cases worldwide and 1,229 in the U.S.

KXAN looked at the U.S. case data for some data points:

Most (777) came from people representing themselves (pro se litigant) Most were in civil courts (1,263), and criminal court cases were in the minority (35) Texas courts caught 79 cases; lawyers filed 36.7% of these KXAN’s approach to AI The frequency of which courts are catching AI errors, or of parties turning to AI tools, has clearly increased over the past years: 11 in 2023, 37 in 2024, 525 in 2025, and 724 in 2026 so far.

“The best way for the courts to catch [AI use] is to do what judges have been doing and verify the authority that the parties are citing,” Appleby said. “I think the most important thing is that the fake cases, although they should not be in any legal briefs that get filed in courts, should not show up in any courtroom because that can be a lot more damaging.”

Penalties can cost thousands While the courts are catching more of the AI-hallucinated cases, it also appears U.S. courts are moving away from imposing monetary fines and instead toward other punishments.

“It depends on the judge and what kind of punishment the judge thinks is warranted,” Appleby said. “And then, I think it depends on how the attorney responds when they realize their mistake. We’ve seen some situations where attorneys have doubled down on this and they’ve attempted to argue that this is real authority and that’s a huge mistake.”

From the logged 2026 cases, courts have struck filings, ordered explanations, issued warnings and referred attorneys for disciplinary action. In some instances, entire cares have been dismissed with prejudice, which prevents them from being reopened.

While most pro se litigants only received warnings for AI errors, some courts also imposed fines.

Monetary fines, for attorneys and pro se litigants, have ranged from $1 and into the tens of thousands. Charlotin’s list consolidates these sanctions with orders to pay legal fees and adverse costs; the largest total penalty was $110,204 in an Oregon case.

It’s on humanity, for now Charlotin’s list can only show a partial outline of AI’s looming shadow over the courts.

Surveys of the legal profession show broad willingness to adopt AI tools; according to legal software company 8am, AI adoption has doubled over the past year and 69% of the profession use AI in their work.

Related: More than half of Texas businesses have adopted AI 8am’s report also notes that while users found efficiency gains, their services aren’t get cheaper. It’s possible that AI spend costs, which will increase as adoption expands, could make legal services more expensive.

Half of respondents said their firms hadn’t provided training about AI use.

“It creates a strong temptation for people to cut corners. It quickly provides confident sound and polished answers to even the most complex questions at times,” Appleby said, “with how much pressure attorneys are under to produce quality work as fast as possible, lawyers will fail to verify what AI gives them and [hallucinations] will end up in court documents.”

AI is powerful but not magic: What AI can and can’t do He also said while he hopes future advances will solve the problem, human verification is currently the only “cure” for hallucinations.

“The attorneys who are signing these papers, they’re subject to ethical rules and obligations no matter who else helped them get them filed,” Appleby said. “What lawyers need to do is to understand … what this technology can and cannot do, and not trust it very much.”