Meta confirmed it will collect employees’ keystrokes, mouse movements, and clicks to train internal AI agents. The company says real‑world user behavior is essential for building tools that help people complete everyday computer tasks. This move follows a Reuters report that the data would be captured via an internal tool.

The strategy taps into a massive data source: Meta’s workforce, far smaller than its 3.5 billion users, but sufficient for training models that mimic human interaction. By harvesting these inputs, the company hopes to reduce reliance on external data sets and accelerate AI development. The policy raises legal and ethical questions about employee surveillance and data ownership.

Critics note that keyloggers in the workplace can violate the Computer Fraud and Abuse Act, and employees have no clear option to opt out or receive compensation. The practice echoes past controversies over AI training data, where companies have faced hefty settlements for using copyrighted content.

Meta’s confirmation, coupled with its silence on worker consent, signals a broader trend of firms leveraging internal labor to fuel AI growth. The decision may prompt regulators to revisit workplace data collection rules.