An employment lawsuit filed in New Jersey this month turned on an unusual witness: the AI assistant in the room.
According to Law360, an environmental scientist at Marathon Engineering has sued her former employer for gender discrimination after an AI assistant recorded her termination meeting and sent her the transcript. That transcript, the suit alleges, captured one of her supervisors saying he hoped to replace her with a “relatively strapping young man.”
Set aside the merits of the case, which a court will sort out. The mechanics are what should get your attention. An AI tool joined a sensitive meeting, recorded everything, and automatically distributed the recording to the one person the company least intended to receive it. Nobody hacked anything. The software did exactly what it was configured to do.
Now picture that same behavior inside a law firm.
Meetings at your firm are not ordinary meetings
AI notetakers have spread quickly because they solve a real problem. They join the video call, transcribe the conversation, summarize the action items, and email everyone afterward. Attorneys like them. Staff like them. Many arrive bundled with video conferencing platforms your firm already licenses, which means they can show up in meetings without anyone deliberately inviting them.
At a law firm, that convenience carries weight that other businesses don’t have to think about:
Privilege. Attorney-client communications stay protected only while they stay confidential. When a third-party vendor stores, processes, or trains on a recording of a privileged conversation, that vendor’s access can hand opposing counsel an argument that privilege was waived. Legal commentators are already predicting that demands for meeting transcripts and AI summaries will become a routine part of discovery.
Consent. Recording laws vary by state, and several states require consent from every participant on the call. A notetaker that auto-joins and starts recording can create liability before the first agenda item.
Ethical obligations. In December, the New York City Bar Association issued Formal Opinion 2025-6 on the use of AI to record, transcribe, and summarize conversations with clients. It advises lawyers to obtain client consent before recording, evaluate where the vendor stores data and how long it’s retained, determine whether the tool trains on that data and whether deletion is possible, and review AI-generated summaries for accuracy before relying on them. The opinion also flags a detail worth sitting with: videoconferencing platforms can make recording the firm-wide default, requiring an affirmative act—often overlooked—to turn it off. And this isn’t slowing down. On August 5, the same committee issued Formal Opinion 2026-2, extending its guidance to recordings of non-client conversations.
Security. Every stored recording and transcript is another repository of client information. Cloud-hosted meeting archives are exactly the kind of concentrated, sensitive data that attackers target — a subject we covered in our post on AI as a security question.
The question that lands on the administrator’s desk
Here’s the uncomfortable part. Most firms cannot currently answer three basic questions:
- Which AI notetakers and transcription tools are currently in use at our firm?
- What have they already recorded?
- Where does that data live, who can access it, and when is it deleted?
The answers exist — inside individual attorneys’ accounts, buried in platform settings, scattered across tools nobody formally approved. This is the same visibility gap we wrote about last week: new research found that firms know who has activated AI tools but rarely know what those tools are actually doing. Assembling the full picture takes time you don’t have, across systems you may not control, involving software you never chose. And when a client questionnaire or an ethics inquiry forces the question, producing the answer becomes an operations problem. Which makes it yours.
You shouldn’t have to chase that down alone. This is a preventable problem, and preventable problems are our job.
Getting ahead of it
Our Managed Intelligence service — part of the Innovative Managed Solution — exists so that firms adopt AI deliberately instead of discovering it after the fact. Applied to meeting AI, that looks like this:
Discovery first. We assess how AI is actually being used at your firm, including the notetakers that arrived uninvited through platform defaults and individual sign-ups. You can’t govern what you can’t see, so we build the inventory before anything else.
Approved tools, configured correctly. We help your firm select note-taking and transcription tools that meet its confidentiality obligations, then configure them so auto-join, auto-record, and auto-distribute behave the way the firm intends — because a tool’s default settings were written for the vendor’s convenience, not your clients’ protection. The Marathon Engineering complaint shows what happens when you “send transcript to all participants” without thinking about it in advance.
A policy people can follow. We develop your AI acceptable use policy, including clear rules for recording: which meetings may be recorded, who must consent, where transcripts are stored, how long they’re kept, and who reviews summaries before anyone relies on them. That gives every attorney and staff member an answer better than guessing. Our earlier post on what new AI guidance asks of your systems covers how these policies connect to your broader infrastructure.
Training that sticks. We train your team on when a notetaker belongs in a meeting and when it absolutely doesn’t, so good judgment doesn’t depend on who happens to remember the policy.
Ongoing review. Vendors change their data practices. Bar guidance keeps evolving. We monitor both and keep your policies current, so a rule written this year doesn’t quietly go stale by next.
All of it sits inside the same managed framework that already handles your firm’s security, monitoring, and risk controls — one partner, one structure, and a straight answer when a client asks how your firm governs AI. If your firm is earlier in its AI journey, our adoption strategy post is a good starting point.
The record is being made either way
AI notetakers aren’t going away, and they shouldn’t. Used well, they save real time and produce better records than memory ever did. The Marathon Engineering case is simply a reminder that these tools create permanent, distributable records of whatever happens in front of them — and that the time to decide who controls those records is before the meeting, not after the transcript lands somewhere it shouldn’t.
We manage the tools, settings, policy, and training, so your attorneys get the productivity they need and your clients get the confidentiality they’re owed. That’s what safe, secure, and reliable means when the technology is sitting in on your meetings.
If you’re not sure what’s recording at your firm right now, schedule a consultation with an Innovative Solutions Consultant, and we’ll help you find out. Request a consultation.
We’ve got your back.
