Meta’s MCI tool logs staff clicks for AI, raising EU privacy alarm

Meta’s Model Capability Initiative is effectively an industrial‑scale data flywheel for agentic AI: record everything employees do on their computers, then train systems to imitate and eventually automate those workflows. From a capability standpoint, it’s a logical extension of RLHF—except the ‘feedback’ is now passive telemetry on real work. If it works, Meta could end up with agents that are unusually competent at software operations across common enterprise tools, an asset that goes far beyond chatbots.([indianexpress.com](https://indianexpress.com/article/technology/tech-news-technology/meta-tool-to-track-employee-mouse-clicks-on-collision-course-with-eu-privacy-rules-10716470/lite/?utm_source=openai))

The flip side is that MCI crystallizes fears about AI development built on pervasive workplace surveillance. Capturing mouse paths, keystrokes and screen contents across 200+ apps inevitably sweeps up personal data, privileged communications and possibly EU citizen information, inviting GDPR scrutiny and labor pushback. If EU regulators decide that this kind of data collection exceeds what’s lawful even with consent, they could force design changes not just at Meta but across an industry that is hungry for human workflow traces.

For the broader AGI race, the outcome will shape whether large firms can safely use internal behavioral telemetry as a differentiating asset. A green light creates a strong incentive for ‘self‑training’ enterprises that treat employees as live curriculum; a red light would push labs back toward synthetic data and more explicit labeling schemes.