What happened
Meta deployed its Model Capability Initiative (MCI) on US employee computers, capturing mouse movements, clicks, keystrokes, and screen snapshots. This data trains AI models, specifically AI agents, to autonomously perform work tasks by improving human-computer interaction, per internal memos and CTO Andrew Bosworth. Meta states the data, part of its Agent Transformation Accelerator (ATA), will not inform performance reviews and includes safeguards for sensitive content.
Why it matters
Employee surveillance for AI training shifts the balance of workplace power, raising significant privacy and ethical concerns for platform engineers and security architects. While Meta asserts data exclusion from performance reviews, potential for misuse creates a new constraint on digital autonomy. This move coincides with Meta's planned 10% global workforce reduction, reflecting a broader industry trend where AI-driven efficiency impacts workforce planning for founders and procurement teams evaluating automation strategies.
