Expert’s Use of GenAI Was Proper on Facts Presented

Ferlito v. Harbor Freight Tools USA, Inc., 2025 WL 1181699 (E.D.N.Y. Apr. 23, 2025), isn’t  the most recent decision concerning an expert’s use of GenAI, but it is an interesting one.

Mr. Ferlito sued Harbor Freight. He alleged that he purchased a “splitting maul,” which is a specialized axe used to split wood, that was defective. While he was hanging the maul up for storage, the head detached and struck him.  He alleged that this was caused by a design defect.  The defense asserted that the cause was misuse.

The decision begins:

“If I had four hours to chop down a tree, I’d spend the first two hours sharpening the axe.”

-Aphorism incorrectly attributed to Abraham Lincoln. See https://quoteinvestigator.com/2014/03/29/sharp-axe/.

One issue turned on the qualifications of the plaintiff’s expert on liability. Another was the reliability of his testimony.

This post looks only at the expert’s use of ChatGPT.  The court found “no issue” with Mr. Lehnert’s use of GenAI, writing:

Federal courts have grappled with the appropriateness of an expert’s use of artificial intelligence to form opinions, and the validity of AI as a research tool in litigation more broadly. See Kohls v. Ellison, No. 24-CV-3754 (LMP)(DLM), 2025 WL 66514, at *5 (D. Minn. Jan. 10, 2025) (excluding expert testimony when the expert’s affidavit contained ChatGPT-generated references to non-existent academic articles)….

In Kohls, the expert’s “citation to fake, AI-generated sources in his declaration … shatter[ed] his credibility with th[e] Court” such that his testimony would not be reliable as required by Rule 702 and Daubert. 2025 WL 66514, at *4. However, the Court emphasized that experts can use “AI for research purposes” given its “potential to revolutionize legal practice for the better.” Id. Daubert issues arise only “when attorneys and experts abdicate their independent judgment and critical thinking skills in favor of ready-made AI-generated answers.” Id. [emphasis added].

The Ferlito court explained:

Here, there is little risk that Lehnert’s use of ChatGPT impaired his judgment regarding proper methods for securing the maul’s head to its handle. The record from the hearing reflects that Lehnert used ChatGPT after he had written his report to confirm his findings, Tr. at 46, which were based on his decades of experience joining dissimilar materials, Tr. at 12. During the hearing, Lehnert professed to being “quite amazed” that the “ChatGPT search confirmed what [he] had already opined.” Tr. at 47. On re-direct, Lehnert reiterated that he did not rely on ChatGPT:

Q: Now, the other items that you have consulted and listed, [including] ChatGPT … those aren’t things that you relied upon in reaching your conclusions and opinions, correct?

A: Correct.

Tr. at 67. There is no indication that Lehnert used ChatGPT to generate a report with false authority or that his use of AI would render his testimony less reliable. Accordingly, the Court finds no issue with Lehnert’s use of ChatGPT in this instance. [emphasis added].

The trial testimony that was cited by the court is not available on Westlaw.  A portion of the expert’s deposition testimony follows:

Q. And under that, you refer to ChatGPT; is that correct?

A.  Correct.

Q. Are these the reference materials that you relied upon on page 7 of your expert report?

A.  Yes. [emphasis added].

Later testimony was:

Q.  Okay. I’d like to go back to what we marked as Defendant’s Exhibit 3, which is that Excel spreadsheet. Again, I want to talk a little bit about the materials that you relied upon or you identified in connection with your report. You identified ChatGPT as a resource that you utilized in connection with your report, correct?

A.  I’ve identified that, but I have to qualify that it was more of a — since I — I wanted to show, and that’s what I’ve done here, what is a collection of common knowledge relative to this subject. And that’s what ChatGPT does, it collects all the information. And I thought it was kind of an interesting exercise that I’d ask at this question. And so when you look at the universe of information that ChatGPT collects and then look at the response that it gives, I thought that very germane to this issue.

Q.  Did you — did — was this the free service of ChatGPT or a paid subscription?

A.  Free service.

Q.  Did you do anything to verify the content of the response that you were given by ChatGPT?

A.  No.

Q.  No independent research to verify the response that ChatGPT gave you?

A.  No.

Q.  Okay. Did ChatGPT provide you any sources where they got their information from?

A.  I don’t recall. I didn’t look for that.

Q.  Okay. You just did a search and this is what came up?

A.  That’s correct.

Q.  Okay. Have you ever used ChatGPT in connection with issuing an expert report previously?

A.  No, I haven’t.

Additionally, the expert stated in deposition: “And my own — my intent on putting all of this reference material together was, I thought very interestingly summed up in that ChatGPT common knowledge.”

A final pretrial conference is scheduled for November 19, 2026.  Docket Entry No. 70 (9/8/2026).

In Her Honor’s discussion of the Ferlito case, K.B. Forrest, New York Practice Series, §79.15. “AI and the courtroom—Federal approach to AI-related evidence and proposed rules” (Oct. 2025), Judge Forrest notes that New York applies the Frye standard and: “By contrast, in Kohls v. Ellison, [2025 WL 66514, at *5 (D. Minn. 2025,] a federal court excluded an expert declaration after learning that the expert had relied on an AI tool to draft portions of the report, including fabricated legal citations. The court held that the expert’s failure to verify the accuracy of the AI’s output fatally undermined the reliability of the testimony under FRE 702, and illustrated the danger of substituting machine output for expert reasoning.”

Judge Forrest wrote:

Taken together, these cases illustrate the emerging contours of permissible AI use under the FRE. Courts may tolerate limited use of generative AI for secondary verification, provided the expert retains independent judgment and applies reliable methods. But experts who defer to AI without scrutiny—or use it in lieu of personal expertise—risk exclusion under FRE 702.

A future post titled “Are an Expert’s A.I. Prompts Discoverable?  Maybe,” will discuss Villanueva v. Las Vegas Metropolitan Police Dept., 2026 WL 2942348 (D. Nev. Sep. 29, 2026).