AI hallucinations help sink multimillion-dollar class action against AIG

Businessman seeking about $34 million from insurance giant withdraws his class action after a judge found five cited rulings did not exist, warning against ‘blind’ AI use and ordering the plaintiff to pay 20,000 shekels in legal costs

A businessman who sought to bring a class action worth an estimated 110 million shekels, about $34 million, against insurer AIG has withdrawn the case after an Israeli judge found that legal filings submitted on his behalf relied in part on nonexistent court rulings apparently generated by artificial intelligence.
The Tel Aviv District Court this week approved the plaintiff’s request to withdraw his motion to certify the case as a class action and ordered him to pay AIG 20,000 shekels in legal costs.
The lawsuit, filed in May 2025, alleged that AIG violated statutory limits on waiting times for customers seeking a human representative by telephone. The plaintiff argued that customers were required to receive a response within six minutes but were in practice waiting an average of 13.8 minutes.
AIG rejected the claim, arguing through attorneys Gad Ticho and Adi Shemesh Balat of Caspi & Co. that the plaintiff had fundamentally misread the applicable rules. The insurer said Capital Market Authority regulations allow a certain share of calls to exceed the target waiting time when compliance is measured over an annual average, and that AIG had met those requirements throughout the relevant period.
The company also argued that the case relied on the plaintiff’s personal experience over only a few months, while the law assesses compliance over a much broader annual period.
Judge Hadas Ovadia said the relevant provisions should have been examined before the certification request was filed.
Without making a final determination on the merits, she wrote that the application appeared to have been prepared “casually,” without examining all the provisions relevant to the claim, and said those provisions would not have supported the plaintiff’s argument had they been properly reviewed.

Five rulings that did not exist

The court also found a more serious problem in one of the legal documents filed by the plaintiff’s attorney: citations to five court decisions that Ovadia said did not exist.
One supposed ruling was cited as supporting the argument that repeated calls by a customer could point to a systemic failure rather than “consumer obsessiveness.” Ovadia said she could find no such ruling.
Another alleged precedent was presented as recognizing that systematic loss of time and inconvenience could amount to compensable damage. Again, the judge said she could not establish that the cited ruling existed.
Ovadia decided, “not without hesitation,” against imposing personal costs on the plaintiff’s attorney payable to the state, but issued a pointed warning about the use of artificial intelligence in legal work.
She stressed that the problem was not the use of AI itself, which she noted some view as a technological shift comparable in importance to the Industrial Revolution, but rather “systematic use that is blind, tendentious and uncontrolled.”
“Human errors can occur in the operation of any machine, including this one,” she wrote, emphasizing the need for lawyers to verify AI-assisted work before submitting it to court.
The judge ultimately approved the withdrawal of the case after recommending that the plaintiff abandon the proceedings.
In her ruling, Ovadia said the action appeared to have been filed in a manner that was “tendentious and misleading,” and could even be viewed as a baseless claim brought for improper leverage. She specifically cited the reliance on AI hallucinations in deciding to award costs to AIG.
The amount was limited, she said, in part because the plaintiff and his attorney agreed to withdraw the case at a relatively early stage.
Ticho and Shemesh Balat, who represented AIG, said the ruling underscored the risks of using AI without professional oversight.
“Any technological tool can be dangerous in unskilled hands,” they said. “Artificial intelligence tools, when used professionally, can improve legal work, as many lawyers already do today. But uncontrolled use can lead to serious errors.”
They added that despite previous court warnings about lawyers’ duty to verify AI-generated material, “serious failures are still being discovered.”
Comments
The commenter agrees to the privacy policy of Ynet News and agrees not to submit comments that violate the terms of use, including incitement, libel and expressions that exceed the accepted norms of freedom of speech.
""