Meta is rolling out tracking software on U.S. employee computers to collect mouse movements, clicks, keystrokes, and some screenshots for training AI tools, according to Reuters. The company says the data will come from a designated list of work apps and is used to build models that can replicate basic computer-use behavior for AI agents. The announcement lands as AI providers compete for “work footage” from real workplace tasks, including OpenAI’s reported outreach to contractors for scrubbed samples of prior work products. For universities and other employers, the Meta disclosure raises immediate questions about privacy controls, consent, and acceptable-use standards for AI training involving human activity data. Meta also positioned the effort as part of its broader agent push, while preparing for potential workforce reductions. The technical promise of agentic systems is colliding with governance expectations around employee monitoring and data handling.