Meta AI training data employee surveillance Mark Zuckerberg privacy

Zuckerberg Defends Employee Keystroke Monitoring in Leaked Audio: 'Smart People Using Computers'

23 May 2026

In a leaked audio recording, Meta CEO Mark Zuckerberg reportedly defended the company's practice of monitoring employees' keystrokes, mouse clicks, and screenshots, arguing the data is needed to train Meta's AI models and give the company a competitive edge over rivals. He claimed the surveillance tool, called the Model Capability Initiative, is not used for performance tracking or employee oversight, but solely to teach AI systems how skilled engineers use computers. Meta is not alone in this approach, as Microsoft and xAI are also reportedly using their own workforces to generate AI training data.

Mark Zuckerberg apparently wants to win the AI race badly enough to turn his own workforce into a training dataset. Leaked audio published by worker advocacy group More Perfect Union captures the Meta CEO delivering a six-minute defence of employee monitoring software during an internal meeting on April 30, the same day Meta announced 8,000 job cuts. Timing, as always, is everything.

In the recording, Zuckerberg reportedly explains that Meta is collecting employee keystrokes, mouse clicks, and screenshots to feed its AI models. The logic: watch smart people use computers, teach the AI to use computers like smart people do. It is not exactly a complex thesis.

"The way that you get a system to be good at using computers is by having it watch really smart people use computers," he purportedly says. "So that's basically the essence of what we are trying to do here."

Meta has not confirmed the audio is genuine, and the company did not respond to press requests for comment. That said, a Meta spokesperson had already confirmed in April that this kind of monitoring was happening. The program goes by the name Model Capability Initiative internally, according to reports.

On the surveillance concerns, Zuckerberg reportedly offered some reassurance. The data is "stripped out" as much as possible, he said, and is not being used for performance assessments or to watch what employees are up to. European staff are apparently exempt from the whole thing, which will surprise nobody given GDPR's fairly clear position on monitoring workers without their explicit consent.

Meta is not alone in raiding its own payroll for training data. The Information reported this week that Microsoft and Elon Musk's xAI are doing similar things with internal teams. Microsoft, sitting on a vast pool of software engineers, reportedly sees its workforce as a structural advantage for improving GitHub Copilot. Same basic idea.

Zuckerberg also drew a contrast between using Meta employees versus external contractors, and was not subtle about it. The average intelligence at Meta, he reportedly said, is "significantly higher" than what you can source through contractor pipelines. The contractors, presumably, were not in the room.

Those contractor pipelines have their own problems. A Wired investigation from January reported that OpenAI's data vendor Handshake AI had been asking freelancers to upload real work products from previous jobs, including financial models, code repositories, and client contracts. OpenAI offered a tool to scrub sensitive details first, but intellectual property lawyers flagged the approach as legally questionable at best.

Zuckerberg's broader framing in the leaked audio is that awkward trade-offs are simply the price of competing at the frontier. Whether Meta's engineers see it quite that way is another matter.