The NLRB Just Made Your AI Meeting Bot a Labor Law Problem
Bindingness: Regulatory Guidance · Scope: Federal · Employment
Part of AI Regulation Tracker, our running brief on the laws reshaping AI at work.
Almost every HR team now runs an AI note-taker in meetings without a second thought. The bot joins, transcribes, summarizes, and files the notes. It feels like pure productivity. The National Labor Relations Board sees it differently. The Board's General Counsel has signaled that intrusive electronic surveillance and algorithmic management can interfere with employees' rights under Section 7 of the National Labor Relations Act, the provision that protects workers acting together over wages, hours, and working conditions. A wave of unfair-labor-practice charges in early 2026 put a real example on the table: AI transcription tools capturing conversations that touched union organizing.
Here is the part HR misses. Section 7 protects most employees, not only those in a union. The right to discuss pay, raise safety concerns together, or talk about organizing is broad, and it does not switch on only when a union is present. So the habit of leaving an AI bot running in every meeting collides with labor law the moment a meeting drifts into wages, working conditions, or organizing, which happens more often than you think. This briefing is about when to switch the bot off and how to write a meeting-notes policy that holds up.
Key takeaways
- The NLRB General Counsel treats intrusive electronic surveillance and automated management as potential interference with Section 7 rights, which protect most employees, not just union members.
- An always-on AI note-taker becomes a labor-law risk the moment a meeting touches wages, working conditions, safety, or organizing.
- The fix is a clear decision rule for when the bot goes off, plus a written policy covering notice, consent, retention, and a sensitive-topic kill switch.
- State recording-consent laws sit on top of this. In two-party-consent states, recording a meeting without everyone's agreement carries its own exposure.
What the NLRB is actually saying
Strip it to the core. The General Counsel's position is that surveillance technologies and automated-management tools can chill protected activity. If employees reasonably believe that everything they say in a meeting is being recorded, transcribed, and analyzed, they may stop speaking freely about the very subjects Section 7 protects, namely pay, conditions, and the possibility of acting together. Tools that monitor, record, or score employees can cross from neutral productivity aid into unlawful interference when they reach protected conversations.
The early 2026 charges made this concrete. An AI transcription tool captured talk related to union organizing, and the recording itself became the basis for an unfair-labor-practice allegation. The Board has also signaled interest in coordinating with other agencies on workplace-monitoring data. The direction of travel is unmistakable: regulators are looking hard at the gap between "we record everything for convenience" and "employees can speak freely about their working conditions."
None of this means AI note-takers are banned. It means they are not neutral. A tool that is perfectly fine in a project status meeting is a liability in a meeting where employees are airing grievances about pay or staffing. The legal exposure does not live in the tool. It lives in leaving the tool on during the wrong conversation.
When HR has to switch the bot off
The practical skill is a decision rule your managers can apply in the moment, not a legal treatise. The line is the subject matter of the meeting, not its title on the calendar.
The bot is fine in a status update and a problem in a grievance. The same tool, two very different conversations. Knowing the difference is the job.
| Meeting touches | Bot status |
|---|---|
| Project status, planning, routine operations | Running is generally fine, with notice and consent |
| Wages, hours, benefits, or pay equity | Switch it off; this is protected territory |
| Working conditions, staffing, or safety complaints | Switch it off; employees may be acting together |
| Anything touching union activity or organizing | Switch it off; recording this is the clearest risk |
| Investigations, discipline, or grievances | Switch it off; sensitive and likely protected |
The simplest version of the rule for a manager to remember: if the conversation is about the work, the bot can stay. If the conversation is about the working conditions, the bot comes off. When in doubt, off is the safe default, because an absent transcript is recoverable through human notes, and an unlawful recording is not.
The policy that backs the rule up
A decision rule needs a written policy behind it, or it dies the first time a busy manager forgets. A defensible AI meeting-notes policy covers four things: clear notice that an AI tool may be used, a consent step that satisfies your strictest recording-consent state, a retention and deletion schedule so transcripts do not pile up indefinitely, and an explicit kill switch instruction for the protected and sensitive topics above. This is exactly the kind of artifact an AI assistant is good at drafting and a human must own.
Claude can draft this policy quickly: give it your states of operation, your current note-taker tool, and the protected-topic list, and it will produce a clear first draft with the notice language, the consent step, the retention schedule, and the kill-switch rule. It moves faster and stays more consistent across locations than writing it by hand. A named human in HR, with employment counsel, then confirms the consent rules against your states' recording laws and signs off before it goes out. For the full policy build, see our companion piece on how HR teams use AI safely.
What to do this quarter
- Inventory where AI note-takers are running today. Most teams have more bots in more meetings than HR realizes.
- Publish the decision rule above so every manager knows the bot comes off for wages, conditions, safety, organizing, and grievances.
- Write a meeting-notes policy covering notice, consent, retention, and the sensitive-topic kill switch, and check the consent step against your strictest recording-consent state.
- Set a default that errs toward off. Human notes can fill a gap; an unlawful recording cannot be undone.
- Train managers on the rule once and reinforce it, because the failure mode is a forgotten bot in the wrong meeting, not a deliberate choice.
Guardrails
A few lines keep this clean. Do not assume Section 7 applies only to union workplaces; it protects most employees acting together over working conditions. Do not rely on a meeting's calendar title to decide whether to record; rely on what the meeting is actually about. Do not ignore state recording-consent law, which in two-party states can create exposure entirely separate from the NLRB issue. And do not let an AI assistant write your consent language and then deploy it without a human confirming it against the specific states where your people sit. Use the tool to draft the policy fast and keep it consistent. Keep the legal call with a person.
What AI does not replace
AI can transcribe your meetings, draft your note-taker policy, and keep the language consistent across locations. It cannot judge in real time whether a given conversation has crossed into protected territory, it cannot decide your state-specific consent obligations, and it cannot carry the liability for an unlawful recording. Those calls belong to a trained manager applying the decision rule and to counsel confirming the consent rules. The assistant makes the policy faster to build and easier to keep current. The judgment about when to switch the bot off stays human. Building that judgment, and the systems around it, is the work we teach in The Leveraged HR Professional.
Frequently asked questions
Does the NLRB issue only apply to unionized workplaces?
No. Section 7 of the National Labor Relations Act protects most employees, union or not, when they act together over wages, hours, and working conditions. An AI note-taker that records employees discussing pay, safety, or staffing can interfere with those rights even if there is no union present. That broad reach is exactly why the always-on note-taker habit is risky.
When should HR turn the AI meeting bot off?
Turn it off whenever a meeting touches wages, hours, benefits, working conditions, safety complaints, investigations, discipline, grievances, or anything related to union activity. The simple rule: if the conversation is about the work, the bot can stay; if it is about the working conditions, the bot comes off. When in doubt, off is the safe default.
Is recording a meeting with an AI tool legal?
It depends on your state and the subject matter. State recording-consent laws apply on top of the NLRB issue. In two-party-consent states, recording a meeting without everyone's agreement carries its own exposure regardless of labor law. A defensible policy covers notice, a consent step set to your strictest state, a retention schedule, and a kill switch for protected and sensitive topics.
Can the NLRB actually penalize an employer for this?
Yes. If the Board finds that recording interfered with Section 7 rights, it can order the employer to stop the practice, post a notice to employees, and provide make-whole remedies, including back pay for anyone disciplined or discharged as a result. The Board does not issue punitive fines, but it can seek enforcement of its orders in federal court. The cost of a forgotten bot in the wrong meeting is real, which is why the decision rule and the policy matter.
Related briefings: The AI employment law patchwork · Illinois HB 3773 AI employment notice · How HR teams use AI safely