Imagine your boss tells you: 'Hey, we're going to watch every keystroke you type and every mouse wiggle you make - for the robots.' Bad enough, right? Now imagine that data promptly leaks internally for anyone at the company to browse. That's the Meta employee experience right now, and honestly, it tracks.

Just two months after Meta rolled out a program to harvest employee keystrokes and mouse movements as AI training data, the company has hit pause - and not by choice. According to a screenshot obtained by Business Insider, private employee conversations, performance data, and transcriptions were floating around and accessible across the organization. Meta confirmed the incident.

How bad was it, exactly?

On Meta's internal severity scale - where 0 is the most catastrophic and 5 is basically a shrug - this was classified as a SEV 2. So not quite 'the building is on fire,' but firmly in 'oh no, this is genuinely bad' territory. The kind of incident that ends careers and starts strongly-worded memos.

The irony is almost too much

Let's sit with this for a second. Meta - a company whose entire business model is built on harvesting human behavior data - decided to start harvesting its own employees' behavior data, then immediately fumbled that data in an internal leak. It's like a locksmith locking their keys in the car. The sheer poetry of it.

The original program was already raising eyebrows. Tracking keystrokes and mouse movements of your workforce to feed an AI system is... a choice. The kind of choice that gets HR departments very sweaty and labor lawyers very excited. But instead of a debate about workplace surveillance ethics, we got an even faster controversy when the collected data stopped being private almost immediately.

What happens now?

Meta says the program is paused. Not cancelled - paused. Which suggests this is less 'we reconsidered our values' and more 'we need to fix the plumbing before we keep collecting the water.' Employees, meanwhile, are presumably left wondering what exactly was in those transcriptions that's now been seen by who-knows-how-many colleagues.

The whole saga is a useful reminder that the race to build better AI comes with some genuinely uncomfortable shortcuts - and that surveilling your own staff is a great way to guarantee that whatever loyalty they had left evaporates entirely. Good luck with those retention numbers, Zuck.