This is what happens when you treat your own employees as unpaid training data.
Meta’s so-called “Model Capability Initiative” was exactly the kind of invasive, overreaching surveillance tactic that makes people rightfully skeptical of Technology. Launched in April, the program tracked employees’ keystrokes, mouse clicks, and basically everything they did on their work computers to feed into the company’s AI models. It logged activity in Gmail, GChat, and even Meta’s own internal AI assistant called Metamate. Screenshot capture was part of the package too.
The logic, according to leaked audio of a company meeting, came straight from CEO Mark Zuckerberg himself. He reportedly said it made sense to use employees because “the average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks.” That’s quite the endorsement of your workforce. Too bad the feeling wasn’t mutual.
More than 1,600 Meta employees signed a petition demanding the program stop. Their message was clear: “Any approach to AI that relies on intrusive, coercive, non-consensual data collection contradicts” the principle of building responsible AI. You can’t really argue with that logic.
Then things got worse. In what can only be described as a catastrophic own goal, the surveillance system exposed private employee data to literally anyone inside the company. According to reporting from Wired, that meant private conversations, AI prompts, transcriptions, and even performance reviews were suddenly accessible to the entire workforce. This wasn’t a glitch in a minor feature. This was the entire point of the program gone spectacularly wrong.
The company has now paused the initiative indefinitely, which is corporate speak for “we got caught and need to figure out what to do next.” The spokesperson’s statement to Wired included the now-standard tech industry defense: “We have carefully designed this program with privacy safeguards.” Clearly, those safeguards needed better design.
Rory Mir from the Electronic Frontier Foundation put it plainly: “Seeking new data for AI training is no excuse” for this kind of worker surveillance. They called it “an abuse of power” and highlighted why we genuinely need legislation protecting worker privacy. When the EFF is telling you you’ve crossed a line, you probably have.
Here’s the uncomfortable part for Meta and the broader Business world: this isn’t an isolated incident. As CNBC reported, nearly every Fortune 500 company is now tracking how much employees use AI tools, trying to measure productivity and ROI. The hunger for AI data is pushing companies to increasingly invasive lengths, and employee monitoring is the obvious path of least resistance when external data sources prove insufficient or too expensive.
Meta is reportedly spending at least $135 billion on AI infrastructure this year, trailing only Amazon, Microsoft, and Alphabet in the spending race. When you’re investing that much money, the temptation to use every available data source becomes almost irresistible. But there’s a fundamental difference between aggressive innovation and treating your own people as guinea pigs without meaningful consent.
The broader pattern here is what should concern everyone. Big tech companies consistently position themselves as champions of user privacy while simultaneously treating their own employees’ digital lives as fair game. If you can’t trust a company to respect the privacy of the people who literally build its products, why would anyone trust them with customer data at scale?
This episode tells us more about where big tech’s priorities actually lie than any public relations campaign ever could.
