Meta is installing keyloggers on US employees' work PCs to harvest mouse movements, clicks, keystrokes, and screenshots. All for training its Muse Spark AI and future models chasing "superintelligence." The company told staff in internal memos that the Model Capability Initiative would only track work-related apps and websites, according to Reuters reporting that surfaced April 21, 2026. Conveniently, Europe gets a pass thanks to stricter surveillance laws.

This isn't some rogue experiment. It's happening alongside plans for more layoffs while Meta pours billions into AI infrastructure. A company spokesperson stressed the data is strictly for model training, not performance reviews, but the optics are brutal: your every hesitation in a menu, every typo in a report, every pattern in how you navigate tools is now raw material to automate someone else's job—possibly yours. X reactions range from "this is just normal workplace monitoring" to engineers calling it the blueprint for replacing entire workflows.

The crunch memo reads like a hostage note from middle management. Meta's aggressive push exposes the contradiction at the heart of Big Tech's AI gold rush: demanding human creativity and precision to build the systems that render humans optional. Employees aren't just workers anymore—they're unlabeled training data walking on two legs. How much is too much? When the tracker is on your machine, the question answers itself.