Meta Monitors Employee Activity to Train AI Agents

The company intends to use these logs to teach AI agents how to interact with digital environments, mimicking human workflows across various applications and websites. Meta CTO Andrew Bosworth outlined the strategy in a recent memo, describing a future where AI handles primary execution while employees shift into roles focused on direction and oversight. The initiative is part of a broader project known as the Agent Transformation Accelerator.

Meta spokesperson Tracy Clayton emphasized that the collected data is strictly for model training rather than performance evaluations. She noted that technical safeguards are in place to mask sensitive content, limiting the scope of captured inputs to specific work-related navigation. Despite these assurances, the move marks a significant expansion of internal data collection as the company pushes to automate standard office operations through its next generation of AI tools.