Meta is rolling out surveillance software called Model Capability Initiative to employees' work computers. The tool records keystrokes, mouse movements, and takes occasional screenshots while workers use applications including Gmail, GChat, VSCode, and an internal app called Metamate. The company says it needs this real-world data to train AI models that better understand how humans interact with computers and complete daily workflows.

CTO Andrew Bosworth reportedly said collecting this data will help Meta realize a vision where AI agents "primarily do the work" while humans "direct, review and help them improve." Meta isn't alone—Anthropic debuted similar technology in 2024, OpenAI announced Operator last year, and Microsoft created cloud PCs specifically for AI agents. The broader industry imagines users delegating tasks like booking flights or scanning shopping sites to AI proxies.

The irony is unmistakable: built on harvesting billions of users' data and frequently running afoul of privacy laws, Meta now applies workplace surveillance to its own staff. CEO Mark Zuckerberg has described this vision as "personal superintelligence" that helps users "achieve goals, create what you want to see in the world, and become the person you aspire to be." Now Meta's remaining employees get a taste of what the company's billions of users have experienced for years.