Skip to content

Menu

LexBlog, Inc. logo
NetworkSub-MenuBrowse by SubjectBrowse by PublisherJoin the NetworkGet StartedSubscribeSupportContact
Search
Close

When your AI meeting assistant becomes your worst witness

By Jon Hyman on August 17, 2026
Email this postTweet this postLike this postShare this post on LinkedIn

“[H]opefully a relatively strapping young man.”

That’s how one manager allegedly described the “ideal person” to replace a female environmental scientist who had just been fired.

That’s bad.

What’s worse (or fortunate, depending on your perspective) was that the company’s AI meeting assistant was still listening.

According to a new sex-discrimination lawsuit against Marathon Engineering, Fireflies recorded the employee’s termination meeting — and kept recording after she left. It then allegedly emailed her a link to the transcript, including the post-termination conversation.

That’s not an AI hallucination. That’s an AI witness.

And it’s a warning for every employer deploying meeting bots and other AI agents.

These tools don’t understand when a sensitive conversation is socially “over.” They record, transcribe, summarize, retain, share, and sometimes automatically distribute information according to their settings.

So, employers, if you’re going to keep using them — and we know you are — they need guardrails around their use.

   1. Keep AI agents out of highly sensitive meetings when possible, including terminations, investigations, discipline, accommodations, and attorney-client communications.

   2. If you use one, stop the recording when the meeting ends. If managers want to debrief, start a new meeting without the bot.

   3. Lock down access. Sensitive transcripts shouldn’t automatically go to every participant, anyone with a link, or an employee you just fired.

   4. Disable automatic distribution for HR matters. Generate first. Review second. Send only if necessary.

   5. Create retention rules. Your AI may be turning conversations that once disappeared into permanent, searchable ESI.

   6. Audit permissions and integrations. Know where the bot goes, what it records, where the information travels, and who can retrieve it.

   7. Train managers to assume the microphone is always hot.

And, it should go without saying, if you’re going to discriminate — just don’t. But if someone is foolish enough to do it anyway, don’t admit it on a recorded meeting.

For years, we employment lawyers have warned managers: Don’t put anything in an email you wouldn’t want blown up on a screen in front of a jury.

AI requires an update: Don’t say anything in front of an AI agent you wouldn’t want copied in a lawsuit or read aloud at your deposition.

Because the next great witness against your company might not be a disgruntled employee.

It might be the AI assistant you invited to the meeting.

     

Related Stories

  • If you don’t know about prompt injection yet, you need to
  • If paid medical leave is good enough for Mitch McConnell, it’s good enough for every American worker
  • Ohio’s new ethics guide — Artificial Intelligence for Lawyers and Judicial Officers — leaves one big question unanswered

 

Tags: AI
  • Posted in:
    Employment & Labor, Technology and AI
  • Blog:
    Ohio Employer Law Blog
  • Organization:
    Jon Hyman
  • Article: View Original Source

Call us at 1-800-913-0988 or email sales@lexblog.com.

Facebook LinkedIn Twitter RSS
The Library at LexBlog
  • About LexBlog
  • The Field We Built
  • Library at LexBlog
  • Our Beliefs
  • Our Team
  • Contact LexBlog
  • Disclaimer
  • Editorial Policy
  • Terms of Service
  • Get Started
  • Publishing Solutions
  • Compass
  • Submit a Request
  • Support Center
  • System Status
Copyright © 2026, LexBlog, Inc. All Rights Reserved.
Law blog design & platform by LexBlog LexBlog Logo