跳到正文
404 Media· Jason Koebler·· 4 小时前AI 评分52

美国司法部用 AI 检测工具 Pangram 质疑辩护律师文书

The Department of Justice Is Using AI Detector Pangram Against Defense Attorneys

AI 导读

美国司法部律师在一起得州选民欺诈案中,用 AI 写作检测服务 Pangram 检测对方辩护律师提交的文书,并将结果作为证据提交,称对方是「defense by ChatGPT」。

正文 · 原文

The Department of Justice Is Using AI Detector Pangram Against Defense Attorneys

Lawyers for the Department of Justice are using the AI writing detection service Pangram to argue that opposing attorneys are using AI to write legal arguments, and in one case told the court that opposing counsel was doing “defense by ChatGPT,” and writing “AI slop” filings, according to court filings reviewed by 404 Media. 

Over the last few years, we have documented many attorneys getting caught by judges using AI after citing hallucinated cases that do not exist. The filings in this court case over the past month show attorneys, including those working for the government, are trying to catch each other using AI, then are citing their opponents’ AI use in their arguments to the court.

“This case has turned into defense by ChatGPT,” government attorneys wrote in the case, which centers around a high-profile voter fraud case. “Motions practice in this case is the equivalent of writing responses to documents generated by a robot that is attempting to practice law. It cannot be that the local rules envisioned that it was acceptable for defense to use what the kids would call ‘AI slop’ to create multiple frivolous motions that waste judicial resources and fail to serve the cause of justice.”

The defendant is Khalwinder Bhengura, an Indian man who entered the United States on an immigrant visa in 2022 and illegally — and apparently accidentally, according to his lawyers — registered to vote in Texas in 2024 after falsely stating that he was an American citizen on his application for a driver’s license. Bhengura was arrested in September on charges of lying about his citizenship status, illegally registering to vote, and illegally voting in the 2024 election, and was found guilty in October on two of the three counts. His family and lawyers have argued that the whole thing is a misunderstanding, and that Bhengura, who is 70 years old and does not speak English well, did not know he was committing a crime and did not understand what he was doing. 

Bhengura has been in prison for months and is set to be deported. He is now not only being held up as an example by the Trump administration of voter fraud but is also now caught in the middle of an AI lawyering dispute.

Bhengura’s case is one of a few dozen in which the Trump administration found disparate cases of voter fraud (which is exceedingly rare) and made big, public shows of the arrests it made. Beyond the political implications of the case, the filings by the Department of Justice show that attorneys are now trying to catch each other using AI, and are centering some of their arguments to the judge around the idea that their opposition is wasting the court’s time with AI-generated filings. 

The case is taking place in federal court in the Northern District of Texas, which recently implemented a rule requiring attorneys to disclose generative AI use in filings. In some of the filings, Bhengura’s attorneys wrote “portions of this brief were prepared with the assistance of a generative artificial intelligence tool.” Government attorneys ran several of the filings through Pangram, the AI writing detection service, and filed the results as exhibits to one of its arguments. 

The Pangram results for the filings were mixed (as in, they did not show that the briefings were 100 percent AI-drafted), but the government’s attorneys wrote that they did not feel like they should have to respond substantively to briefs that were partially AI-generated.

The United States has “been trying to economize the amount of effort it puts into responding to Defendant’s meritless motions given that Defendant appears to be almost entirely using AI to generate them,” the DOJ attorneys wrote.

“As the Court can see from a cursory review of the results of Bhengura’s motions, most of the non-AI generated text in these filings consists of statements from third parties other than Defense counsel, including statements from evidence or documents in the record or quotes from caselaw. In other words, virtually nothing of substance in these filings seems to have been drafted by a human on the Defense team,” they added.

The Department of Justice Is Using AI Detector Pangram Against Defense Attorneys
The Department of Justice Is Using AI Detector Pangram Against Defense Attorneys

The judge in the case, Mark Pittman, chastised the defense attorneys for their use of AI, and wrote, “This Court is disturbed by counsel’s repeated filings which subtly bend the law by proffering superficially persuasive but fundamentally meritless arguments so clearly generated by artificial intelligence. Defendant is not a pro se litigant; he has three counsel of record with decades of practice between them. It is distressing to see how quickly the advent of artificial intelligence has undermined counsels’ ability to properly advocate for their client.”

Nicole Knox and Mario Herrera, two of Bhengura’s attorneys, told 404 Media that the story isn’t so simple, and say they felt they had to use AI because of a health emergency and in order to meet the court’s tight filing deadlines ahead of the election. Herrera said Knox “experienced a medical emergency while we were at the U.S. Attorney's Office preparing the case with the prosecutors […] thus, I prepared a federal jury trial in four days with the lead attorney in the hospital.”

“I represented my client to the best of my abilities within this limited timeframe. I utilized all available legal tools and did not violate any rules or laws in doing so,” he said. 

“The government's ‘evidence’ consisted of output from Pangram, an AI-detection software. An AI-detection software identifies patterns in writing styles based upon a document’s content. It cannot determine who wrote a document, how it was written, or if the law applied within a document is correct,” Herrera added. “The U.S. Attorney's Office did not identify any false citations. ‘Defense by ChatGPT’ is a label the government placed in a court filing regarding the defense of a 70-year-old permanent resident facing prison and deportation. It is not a factual statement. And the attorneys who wrote it were present when Ms. Knox collapsed. They knew precisely why I was standing alone with a new co-counsel in that courtroom four days later. There was nothing requiring them to state what they stated.” 

Pangram CEO Max Spero told 404 Media that anecdotally, he has heard that law firms have begun to use his company’s software, but that most use cases he has seen have been from firms using the tool internally to check whether their junior attorneys and paralegals are using AI to draft legal briefs. 

Herrera said that the fact that Bhengura was found not guilty on one of the three counts he faced is evidence that “Mr. Bhengura received the representation that he deserved.”

“Lawyers have been using templates, forms, and drafting software for decades. This is merely a newer version of those very tools. The arguments that the judge attacked were written by me. Some portions of the filings were written with the aid of AI but then reviewed and edited by me. All final work product was mine and I stand behind the argument and logic of each and every document,” he added. “It was the use of AI during that week that allowed me to work on many tasks at once and produce so much work in four days.” 

Knox added, “AI could have never come up with our argument,” and said that the new AI disclosure rule has made it such that “if you obtain one sentence or phrase through the use of A.I., then you must disclose it. We have nothing to hide.”

“Mr. Bhengura is 70 years old, a lawful permanent resident of this country, speaks Punjabi, and his English is limited. Our defense on the one count of which the jury acquitted him was that he never knowingly claimed to be a U.S. citizen,” Herrera said. “He has been in custody since the verdict, and in addition to whatever sentence he receives for the offenses of which he has been found guilty, he faces deportation from this country. He is entitled to more compassion and humanity from everyone involved in this case than he has received so far.”

The Department of Justice did not respond to a request for comment.

来源:404 Media · 404media.co