Meta widens employee tracking to train AI, raising privacy concerns

Meta Platforms has turned its own workforce into a training set. The company installed tracking software on U.S.-based employees’ computers to collect mouse movements, clicks and keystrokes for artificial intelligence development, a system internally called the Model Capability Initiative, or MCI. The tools also monitored how staff used popular services such as Google, LinkedIn and Wikipedia, and later reporting said the system could capture periodic screenshots and other screen activity on work devices.

Meta said the purpose was to train AI agents that could eventually perform office tasks autonomously. But the same data that helps models learn how people navigate software can also expose what workers search for, how they communicate and where they spend their time, which has made the project a flash point inside the company and beyond it. Employee backlash has included internal protest activity and flyers, while some workers have objected to what they see as surveillance-like monitoring in the name of innovation.

The sharpest legal fault line is in Europe. Meta has said European employees were excluded from the program because the General Data Protection Regulation would not allow equivalent monitoring. That distinction matters because the project is not just about one company’s internal practices. Reuters-linked reporting said the system may also capture non-U.S. data in the process, raising questions about cross-border data flows, consent and whether behavioral data collected in one jurisdiction can be repurposed for AI training elsewhere.

That tension lands in a region where regulators have already pressed Meta on AI training. In June 2024, Ireland’s Data Protection Commission said it welcomed Meta’s decision to pause plans to train a large language model using public content shared by adults on Facebook and Instagram across the EU and EEA after regulatory engagement. In December 2024, the European Data Protection Board issued Opinion 28/2024 on data protection issues tied to AI models, reinforcing the idea that GDPR principles and AI development must be reconciled case by case rather than treated as separate goals.

The timing has sharpened the fight. Reuters reported in May 2026 that Meta was cutting about 8,000 jobs, roughly 10% of its workforce, as it pushed harder toward AI. For workers, that has made MCI look less like a neutral productivity experiment and more like a preview of how the next generation of workplace AI could be built: with intimate data from employees who may have little say in how it is collected, analyzed or reused.