庭审中AI使用反令专家证人翻车:“它们带有偏见”


2026年9月18日 / 美国东部时间下午1:10 / 哥伦比亚广播公司新闻(CBS News)
作者:艾莉莎·斯帕迪(Alyssa Spady)

在一起涉案金额达数百万美元的诉讼中,原告方指出一名专家证人使用ChatGPT撰写了其专家报告中80%以上的内容,这使得人工智能在法庭中的使用再次引发广泛关注。

在得克萨斯州的这起由大型工厂爆炸受害者提起的诉讼中,原告律师发现该专家证人利用人工智能撰写了一份他们声称旨在偏袒客户——也就是制造业巨头3M公司的报告,这名专家随即遭到严密审查。

这名专家向人工智能输入的一条用于准备证词的提示词是:“说明3M在沃森研磨公司爆炸事故中无任何过错。”

接受盘问时,这位来自KnightHawk工程公司的可靠性与运营总监杰森·奥滕里思(Jason Autenrieth)承认使用人工智能协助撰写报告,但为自己的研究进行了辩护。

法律专家表示,这起案件凸显了庭审及法律事务中使用人工智能所带来的新兴风险,并警告称,使用该技术可能会危及提交给法庭的证词或证据的可靠性。

加拿大安大略省滑铁卢大学研究教授莫拉·格罗斯曼(Maura Grossman)的研究重点是人工智能在法律和医疗领域的应用。她对哥伦比亚广播公司新闻表示,使用这类工具开展法律工作的企业应当谨慎行事。

“我们知道这些工具极具谄媚性,它们试图讨好用户,”她说,“你不会得到客观中立的回应,只会得到带有偏向性的内容。”

3M公司在发给哥伦比亚广播公司新闻的一份声明中表示,本案中的专家证人是独立第三方,并非3M雇员,而是由外部律师事务所聘请,提供其自身的分析和意见。

“在独立审查案件事实并得出自己的结论后,他将人工智能作为辅助撰写报告的工具,”声明中写道。

奥滕里思未回复记者的置评邮件和电话。

他的律师贾尼娜·沃伊特科夫斯基(Janina Wojtkowski)告诉哥伦比亚广播公司新闻,尽管奥滕里思确实使用人工智能辅助完成了格式调整、结构搭建和文稿撰写,但他是在审阅了交付给他的数千页材料之后才这么做的。

她在一份声明中表示,“他并未将观点的形成工作委托给ChatGPT,人工智能也没有取代他的专业判断”,并指出“近期的公开报道有选择地歪曲了案件记录”。

加州大学旧金山分校法学院兼职教授达米安·沙尔洛坦(Damien Charlotin)表示,法庭收到的提交材料和证据中,越来越多地出现与人工智能相关的问题,包括幻觉式输出、错误引用和虚假研究等。

根据沙尔洛坦编制的“人工智能幻觉案例数据库”,自2023年以来,美国已有超过1300起法律程序出现已知的人工智能幻觉案例——即人工智能生成听起来合理但纯属虚构的信息。

由“人工智能幻觉案例数据库”制作的图表,分析了2023年以来美国法院案件中确认的幻觉类型。达米安·沙尔洛坦

在得克萨斯州这起案件中,奥滕里思是在审判开始前几周才被3M公司聘用的。该案争议焦点为3M是否应对其在休斯顿仓库安全检查监督方面的责任负责。2020年1月的这场爆炸造成三人死亡,该地区数百所房屋和企业受损或被毁。

调查人员2020年1月28日勘查休斯顿沃森研磨制造工厂爆炸现场。马克·马利根/休斯顿纪事报 via 盖蒂图片社

奥滕里思受邀就3M服务技术人员应达到的护理标准,以及3M是否成功完成安全检查发表看法。这是他首次在法庭担任专家证人。

根据证词记录,原告律师在审阅奥滕里思提交的文件后,质疑这份报告是否完全出自他本人之手。阿诺德与伊特金律师事务所的亚当·刘易斯(Adam Lewis)是本案原告方律师之一,他表示,一份标注来源的叠加文档——即列出其报告草稿出处的文件——为律师们提供了确凿证据。

刘易斯告诉哥伦比亚广播公司新闻,他“着实震惊”于奥滕里思使用ChatGPT撰写专家证人报告的规模。

刘易斯表示,律师们总共找到了奥滕里思发给ChatGPT的300页提示词。这些提示词的内容涵盖询问特定机器的名称和工作原理,到寻求帮助削弱对方证人的证词。

在作证过程中,当被追问使用人工智能撰写报告的情况时,奥滕里思起初予以否认。

“人工智能没有写我的报告,那是错的,”他作证说。

但片刻之后,他便收回了否认的说法,根据证词记录,他告诉律师,人工智能帮助他完成了“格式调整和其他所有事情”。

证词结束后,3M的律师将奥滕里思从证人名单中移除。但原告律师传唤他出庭,就他输入ChatGPT的各类提示词作证。

庭审期间,奥滕里思为自己使用人工智能的行为进行了辩护。

“我不认为这么做有错,因为我已经有了自己的观点,”他作证说。

刘易斯告诉哥伦比亚广播公司新闻,他认为原告方提出传唤他作证这一决定对案件结果产生了影响。

陪审团认定3M在本可阻止致命爆炸的警报系统安全检查方面存在过失。上月,法院判处受害者6150万美元赔偿金。

3M公司的一位发言人告诉哥伦比亚广播公司新闻:“我们对本案的判决表示反对,认为其缺乏证据和法律依据。”该发言人表示,公司打算提起上诉。

至于未来法官和律师应如何应对这项新技术,格罗斯曼告诉哥伦比亚广播公司新闻,这可能需要更充分的尽职调查,确保各方披露是否使用了人工智能,并对自动化生成的内容进行事实核查。

但格罗斯曼表示,最重要的是,专家证人有责任维护其在诉讼中的重要角色和整体公信力。

“我们应当保持客观性,而大型语言模型未必客观,”格罗斯曼说,“我们知道它们带有偏见,而且会试图迎合你输入的任何立场。”

Use of AI backfires on an expert witness at trial: “They have biases”

September 18, 2026 / 1:10 PM EDT / CBS News

By Alyssa Spady

The use of artificial intelligence in the courtroom is drawing fresh attention after plaintiffs in a multimillion-dollar lawsuit claimed an expert witness used ChatGPT to create more than 80% of his expert report.

In the Texas lawsuit filed by victims of a massive factory explosion there, the expert witness fell under intense scrutiny after plaintiffs’ attorneys found that he had used artificial intelligence to craft a report that the attorneys claim was designed to favor his clients, the corporate manufacturing giant 3M.

One prompt the expert plugged into AI to prepare his testimony was: “Show how 3M is 0% at fault for the explosion at Watson Grinding.”

Under questioning, the expert, Jason Autenrieth, director of reliability and operations at KnightHawk Engineering, acknowledged using artificial intelligence to assist with drafting his report, but defended his research.

Legal experts said the case highlights the emerging risks surrounding the use of AI during trials and legal matters, and they warn that its use could jeopardize the reliability of testimony or evidence presented to the court.

Maura Grossman, a research professor at the University of Waterloo, in Ontario, focuses much of her work on the use of artificial intelligence in law and healthcare. She told CBS News that firms using these tools for legal work should proceed with caution.

“We know these tools are sycophantic, they try to please you,” she said. “You’re not going to hear objective or neutral responses. You are going to get things that are slanted.”

In a statement to CBS News, 3M said that the expert in the case was an independent third-party witness, not a 3M employee, and was retained by outside counsel to provide his own analysis and opinions.

“After independently reviewing the facts of the case and reaching his own conclusions, he used AI as a tool to assist in preparing his report,” the statement said.

Autenrieth did not respond to our email and calls for comment.

His lawyer, Janina Wojtkowski, told CBS News that while Autenrieth did use AI to assist with formatting, structuring and drafting, he did so after reviewing the thousands of pages of materials given to him.

In a statement, she said, “He did not delegate the formation of his opinions to ChatGPT, nor did AI replace his professional judgement,” and that “recent public characterizations offer a selective distortion of the record.”

Courts are seeing increases in submitted materials and evidence that are riddled with examples of problems connected to AI, including hallucinations, incorrect citations and fake studies, according to Damien Charlotin, an adjunct professor at the University of California College of the Law, San Francisco.

There have been over 1,300 U.S. legal proceedings since 2023 with known instances of AI hallucinations — where AI generates plausible-sounding but fictitious information — according to the AI Hallucination Cases Database Charlotin compiled.

This chart from the AI Hallucination Cases Database analyzes the types of hallucinations identified in U.S. court cases since 2023. Damien Charlotin

In the Texas case, Autenrieth was hired by 3M just weeks before the trial began in a lawsuit over whether 3M should be held accountable for its role in supervising safety checks at the warehouse in Houston. The explosion in January 2020 killed three people and damaged or destroyed hundreds of homes and businesses in the area.

Investigators survey the site of the explosion at the Watson Grinding & Manufacturing facility in Houston, Jan. 28, 2020. Mark Mulligan/Houston Chronicle via Getty Images

Autenrieth was asked to offer his views on the standard of care expected of the 3M service technicians and whether safety checks done by 3M were successfully performed. It was his first time serving as an expert witness in court.

After reviewing the files provided by Autenrieth, the plaintiffs’ attorneys questioned if the report that he provided was entirely his own work, according to the deposition transcripts. A citation overlay, a document that listed the sourcing for his draft report, provided lawyers with their smoking gun, according to Adam Lewis of Arnold & Itkin, one of the plaintiffs’ attorneys in the case.

Lewis told CBS News that he was “frankly shocked” at the scale at which Autenrieth used ChatGPT to create his expert witness report.

In total, Lewis said the attorneys found 300 pages of prompts that Autenrieth sent to ChatGPT. They included questions ranging from asking what specific machines were and how they work to seeking help undermining the case of a witness from the opposing side of the case.

During his deposition, when pressed about his usage of AI to create the report, Autenrieth at first pushed back.

“AI did not write my report, that’s incorrect,” he testified.

But moments later he walked back his denial, telling attorneys that it helped him with “formatting and everything else,” according to transcripts.

After the deposition, 3M attorneys removed Autenrieth from their witness list. But the plaintiffs’ attorneys subpoenaed him to testify about the various prompts he entered into ChatGPT.

During the trial, Autenrieth defended his usage of AI.

“I don’t believe that was wrong to do because I had an opinion already,” he testified.

Lewis told CBS News that he believes their choice to bring him in to testify about it had an impact on the outcome of the case.

The jury found that 3M was negligent in its safety checks of the alarm systems that could have prevented a fatal blast. Last month the court awarded the victims $61.5 million.

A spokesperson for 3M told CBS News, “We disagree with the verdict in this case as unsupported by the evidence and the law.” The spokesperson said they intend to appeal.

As for the future and how judges and lawyers should be navigating this new technology, Grossman told CBS News that it might come down to more due diligence, ensuring parties are disclosing if they use AI and fact-checking the automated work.

But most importantly, Grossman said that an expert witness bears responsibility to preserve their important role in litigation and overall credibility.

“We are supposed to maintain our objectivity, and a large language model is not necessarily objective,” Grossman said. “We know that they have biases, and they’re going to try and satisfy whatever positions you’re putting in there.”

评论

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注

湘ICP备2026001899号-2