Virtual Termination Meetings And AI: When Skynet Does Not Stop

What can go wrong if an employer forgets to shut off its AI transcriber after a virtual termination meeting ends? One New Jersey employer may find out soon. In the “You Can’t Make This Up” category, a New Jersey plaintiff sued her former employer on August 12, 2026, alleging she was the subject of gender discrimination after an AI-generated transcript summarizing her virtual termination conference allegedly reported that a supervisor said the “ideal person” to replace plaintiff was “[h]opefully a relatively strapping young man.” This set of unusual facts may be a harbinger of things to come as artificial intelligence in the workplace becomes more common. Whether the AI-generated account will be admissible in court remains to be seen. Still, the lawsuit presents an immediate lesson for employers: AI tools can preserve and potentially disclose conversations participants never intended employees, opposing counsel, or courts to see. 

The Lawsuit

According to her lawsuit, plaintiff Lindsay Waninger worked for Marathon Engineering for fewer than three months when she was fired in a video conference that was transcribed using the “Fireflies” AI app allegedly set up by Marathon. Waninger subsequently received a link to the transcript of the termination meeting from Fireflies.ai. The transcript allegedly included the discussion after plaintiff dropped off the call. During that private internal discussion, the supervisors allegedly spoke about “what our ideal person looks like,” to which one supervisor allegedly replied that a candidate to replace plaintiff was “[h]opefully a relatively strapping young man.”

Is An AI-Generated Summary Evidentiary?

One potentially significant issue in the Waninger v. Marathon Engineering & Environmental Services Inc. litigation will be whether, and for what purposes, a non-verbatim AI-generated summary is admissible in court. Given the novelty of AI-generated evidence, the courts of New Jersey (and most other states) have not weighed in.

Recognizing the power and opportunity AI tools can offer but also the very real prospect that generative AI systems can produce inaccurate or fabricated answers, commonly referred to as hallucinations, New York courts emphasize the need to test AI-generated evidence under the approach set out in Frye v. United States, 293 F. 1013 (D.C. Cir. 1923). The Frye approach asks whether the accepted techniques, when properly performed, generate results accepted as reliable within the scientific community generally.

This approach was adopted by New York’s highest court in People v. Wakefield, 38 N.Y.3d 367, 381 (2022), which affirmed the need for a Frye hearing in a criminal proceeding to establish that the relevant scientific community generally accepted the use of proprietary software to interpret DNA evidence before that evidence was admitted. That hearing included expert testimony explaining the mathematical formulas, the processes involved, and the peer-reviewed articles published in scientific journals.

More recently, in Matter of Weber, 220 N.Y.S.3d 620 (Sur. Ct. 2024), a lower court also applied Frye when faced with a valuation expert who relied on Microsoft Copilot to cross-check his calculations. The expert was rejected, in part, because he could not recall what input or prompt he used to assist him with his damages report; could not state what sources Copilot relied upon; and could not explain how Copilot works or how it arrives at a given output. Moreover, the court entered a similar prompt into Copilot three times and came up with three different results—suggesting these variations call into question the reliability and accuracy of Copilot to generate evidence to be relied upon in a court proceeding. Interestingly, when the court asked Copilot: “Are you accurate?” Copilot’s AI-generated answer was: “I aim to be accurate within the data I’ve been trained on and the information I can find for you. That said, my accuracy is only as good as my sources so for critical matters, it’s always wise to verify.” Id. at 634.

The Future Is Now for New Jersey and Beyond

Artificial intelligence, machine learning, natural language processing, and AI assistive materials permeate the modern workplace and are here to stay. Due to the rapid evolution of artificial intelligence and its inherent reliability issues, any evidence generated by an AI product or system should be subject to a Frye hearing prior to its admission. That should provide an objective, unbiased review of the reliability of the purported evidence.

As for the reliability of the Fireflies app, and myriad other AI transcription applications available for Zoom, Teams, and the like, a Frye hearing will be helpful for all parties to assess the reliability of the app generally and how it relates to its specific use by the employer. Courts are unlikely to unquestioningly accept as accurate anything performed by artificial intelligence. However, even if the transcript is ultimately inadmissible in court, its existence likely will need to be disclosed during discovery. It could identify additional witnesses, prompt additional investigation, and possibly create a damaging factual dispute that might not have arisen otherwise.

The Bottom Line

From cell phone voice recorders to Meta glasses to AI transcribers, employees have more ways than ever to document employer actions. Employers should use great care when interacting with employees over any virtual platform, especially when an adverse action is contemplated. It might be old school, but in-person termination meetings should be considered going forward. Whether a sensitive meeting occurs in person or virtually, employers should assume that recording and transcription technology may be present and manage the conversation accordingly.

Employers who do continue to use virtual meetings for employment purposes may want to consider these additional steps to protect themselves:

• • Establish rules governing when AI transcription and meeting summary tools may be used, particularly during disciplinary, investigation, accommodation, and termination discussions.
  • Designate an individual to confirm that recording/transcription has stopped before participants begin any post-meeting discussion.
  • Ensure participants understand whether and to whom the platform will automatically distribute transcripts, summaries, recordings, or links. 
  • Ensure participants understand that an AI platform will not automatically stop recording a meeting merely because a participant has left the meeting.
  • Assume statements made during or immediately after a virtual employment meeting may someday be read by the employee’s lawyer.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.