Meta just paused its AI training program. 1,600 employees revolted. Your employee surveillance plan just got a case study.

Meta quietly built a program that recorded every keystroke, mouse click, and screen capture from its employees' laptops.

The goal: train AI models on how "really smart people" do their jobs.

Then something broke.

Not the technology. The trust.

1,600 Meta employees signed a petition demanding the program be cancelled. They couldn't opt out. They couldn't see what was being collected. And nobody could tell them whether privacy reviews had actually been completed.

Meta's response was a 30-minute pause button. Thirty minutes of not being watched. Per day.

Then it got worse.

Internal security documents revealed the collected data — full prompts, transcriptions, private conversations, performance reviews — was accessible to anyone inside the company. Wired and Business Insider confirmed it was classified as a SEV-2 security incident.

Meta paused the entire program on June 22.

Here's what your workforce team needs to understand.

Zuckerberg told employees that AI improves by watching smart people work. That logic is spreading. Every company building AI on employee data is facing the same question Meta just answered badly: how do you collect enough to train models without destroying the trust that makes people productive?

The answer is not "collect everything and ask later."

If you're building internal AI that learns from employee workflows, your first move is a transparent data governance framework with opt-in consent, defined retention periods, and clear boundaries on what gets used for training versus what stays in HR systems.

Meta had 1,600 employees tell them what that looks like. Most companies won't get a petition. They'll get resignations.