Two years ago, the owners of a greyish, two-story house near John F. Kennedy International Airport decided to sue their tenant. The landlords, a married couple, claimed the 80-year-old woman living there owed them more than $40,000 in unpaid rent, and hired an attorney to get it back.
Instead, their lawyer used fake, likely AI-generated court cases to support his argument that the lawsuit should move forward, according to Queens Housing Court Judge Kimon Thermos's decision to sanction the lawyer.
Attorney Innocent Chinweze filed an argument to the court in April that included seven "nonexistent cases," Thermos wrote. When the judge confronted Chinweze months later in a June hearing, Chinweze claimed he was "hacked," and sent the judge 88 pages of additional research that was also likely generated by AI, according to court records.
Thermos fined Chinweze $1,000 on July 29 for submitting the fictional cases and referred him to one of New York's attorney grievance committees—a state court group that could suspend or revoke his license. In his decision, Thermos wrote that Chinweze had "undermined his own credibility" by detailing the "alleged hacking," but then declining to bring witnesses to prove that it happened.
Chinweze admitted to the court (and to Hell Gate) that he used Microsoft Copilot for research—Copilot is integrated into his laptop, Chinweze said. But he maintained that it was a hack, not Copilot, that caused the fictional cases to end up in his legal filing.
"Because my computer had already been contaminated, jumbled cases that I never knew [and] have never seen were there," Chinweze said over the phone. "I only became aware of it when the judge entered the final judgement on the case."
Chinweze, oddly enough, agrees with the judge's decision to fine him. Even though he claims that the error was "beyond [his] control," Chinweze said, he "should have caught it." He said he planned to pay the fine, and that he has "no problem with that."
"It was a major error, but since it came from me, I have no choice but to accept that responsibility," Chinweze said. Judge Thermos, he added, "was right. He did what he needed to do to protect not just the integrity of the legal profession, but also the integrity of the people that we serve."
Chinweze was representing landlords Felix Ogieva and Happiness Idehen in their court case against their tenant Gloria Stoute-Phillip, who says she's lived in the Queens building since the 1980s, according to court documents. (Ogieva and Idehen told Hell Gate they have found a new lawyer, and plan to file a complaint against Chinweze with the New York State Bar Association.)
The tenant's lawyer, Legal Aid Society Staff Attorney Stephanie Ramdhari, said she had asked the judge to put the back rent case on hold while a separate discrimination case brought by Stoute-Phillip against the owners worked its way through Manhattan Supreme Court. (Stoute-Phillip claims that the landlords discriminated against her by refusing to accept her housing voucher; the landlords deny that allegation.) Chinweze argued that the case should be brought to trial, and referred to 11 cases to explain why.
But when Judge Thermos poked into Chinweze's legal arguments, he found that seven of those cases were extremely fishy. Some of them had fake names and real case citations—a reference number that lets lawyers and judges look up individual cases. Others had real names but invalid citations. And still others had fake names and fake citations.
In one example, Thermos looked up the citation of a case with a fake name, and found that it actually referred to a real medical malpractice case—something that has absolutely nothing to do with landlord and tenant law, the judge wrote in May. (And even if those fake cases "actually existed," they were all more than a decade old "and do not reflect the current state of the law," Thermos added.)
Thermos decided to dismiss the back rent case, and ordered Chinweze to explain why he shouldn't be sanctioned for citing cases that do not exist. Chinweze offered a 94-page response on June 2, admitting that he used Copilot, and "did not know and could not contemplate" that it could generate fake cases. But he said that "malware and unauthorized remote access" had "contributed to the generation or alteration" of the cases he cited, and he failed to notice the mistakes because of a serious health issue he's had since the beginning of the year. (Chinweze provided the same rationale to Hell Gate.)
He also attached an 88-page appendix that he told Hell Gate was a "manual review" of the cases "relevant to the issue." But that appendix, Thermos discovered, was also partially generated by AI.
A screenshot of part of Chinweze's appendix. (Hell Gate)
Three case summaries Chinweze provided included the lines: "Some case metadata and case summaries were written with the help of AI, which can produce inaccuracies. You should read the full case before relying on it for legal research purposes." And most of the cases Chinweze cited weren't relevant at all, Thermos added.
The appendix lists seemingly random cases, including one where a former NFL player sued a company he hired for fraud and another where a Sanitation worker sued the City for negligence—two things that had nothing to do with the lawsuit at hand, Thermos wrote in his decision to discipline Chinweze. It "inexplicably" references the rate of interest on a money judgment, "but no explanation is provided as [to] why that would be relevant," Thermos wrote.
Thermos called it "an incoherent document" that "has no structure, contains the full text of most of the cases cited, shows distinct indications that parts of the discussion/analysis of the cited cases were written by artificial intelligence, cites multiple cases that do not have anything to do with the issues presented…and cites to multiple cases for propositions that are not remotely supported by the case cited."
Chinweze's appendix abruptly includes general information about housing court. (Hell Gate)
As for the alleged hack, Chinweze told Hell Gate that staffers from a Chicago-based IT company he routinely works with examined his laptop and confirmed that he had been hacked. But he said it would be too expensive to pay the IT workers to explain what happened to the judge, because he was only making "about $1,500" on the case. (He also declined to provide the name or contact information for the IT company, saying they "shy away from court and shy away from news.")
"You can blame Copilot. It's still my responsibility," Chinweze said. "I will never allow myself to make this mistake again."
He added, "Nobody was affected adversely, because the court picked it up, and the court asked me, and I explained myself."
Thermos did catch those fictional cases, and the misuse of AI in New York's courts is rare, the judge wrote. (A quick search of state court decisions revealed no other housing court cases involving artificial intelligence.) Ramdhari, the tenant's lawyer, told Hell Gate she had never encountered this issue until this case.
But Ramdhari worries that if more attorneys begin relying on AI-generated case summaries and appear before a less attentive judge, the consequences could be dire for tenants in housing court.
"It's very likely that an unrepresented party is not going to know that the other side's argument is based upon an artificially generated case that fits their narrative," Ramdhari said. "Particularly, vulnerable tenants might just concede to whatever the other side is telling them, and waive rights that they might have, or eventually lose their own housing based on a misrepresentation."
Nationwide, plenty of lawyers have been sanctioned for relying on AI-hallucinated case law to back up their claims. And it's a trend that's only likely to grow, said Sateesh Nori, a former housing court lawyer who now builds AI tools for tenants.
"What's troubling is it's probably easier now than ever before to come up with stuff that sounds credible, but is not real," Nori said. "Judges are going to have to be even more vigilant."
AI has also become a growing technology for New York City businesses and government, even though it can make incredibly embarrassing errors. The City has also used AI to impersonate Mayor Eric Adams's voice and to create a chatbot that can answer questions from small business owners—who the chatbot then told to break the law.
Alex Jacobs, a staff attorney with Queens Legal Services' Housing Unit, said that reliance on AI tools could cause uninformed attorneys to submit false cases on accident—with devastating consequences.
"There can be people that never practice in this area, who might not be particularly savvy and might not understand that an AI is not necessarily producing accurate results, and there's a very real risk that those individuals can be submitting papers to court to make somebody homeless where there's no legal basis for it," Jacobs said.