ai training data / employee monitoring

Meta will begin collecting employees' mouse movements, keystrokes and other desktop activity to build internal AI models, according to reports. The company says data capture is intended to improve productivity tools and AI assistants, but raises privacy and security concerns because the logs could include sensitive information and personal data. Meta plans to combine activity telemetry with other internal datasets to train models that understand workflows, which may accelerate development of enterprise-facing AI features. Employees and privacy advocates worry about scope, consent, retention and guarding secrets, while engineers note potential benefits for tooling. The move underscores tensions between AI training needs and workplace privacy governance.

Meta is collecting employee mouse movements and keystrokes to build datasets for training internal AI models, Reuters reports. The company says the data will be used only for model training and not for performance evaluations, drawing skepticism from staff amid ongoing layoffs and cost cuts. The move matters because employee interaction logs can produce high-quality, real-world training signals for productivity and agent-style models, accelerating internal AI development while raising privacy and trust concerns. If broadly applied, this practice could shape workplace AI behavior, influence future tooling, and spark debates over consent, surveillance, and data governance in tech firms.

Meta is rolling out tracking software on U.S. employees’ work computers to record mouse movements, clicks, keystrokes and occasional screen snapshots to train AI models that can perform workplace tasks autonomously. Internal memos shared by staffers say the tool will monitor interactions on a specified list of work apps and sites to improve model performance on UI tasks like dropdown selection and keyboard shortcuts. Meta spokesperson Andy Stone said the data will be used solely for model training, not performance reviews, and that safeguards are in place to protect sensitive content. The move highlights how companies are sourcing granular behavioral data to build practical workplace AI, raising privacy and security considerations.

Meta will deploy tracking software on U.S. employees’ work computers to log mouse movements, clicks, keystrokes and occasional screen snapshots for training AI models, Reuters reports. The initiative targets interactions within specified work apps and websites to improve agents’ abilities on tasks like selecting dropdowns and keyboard shortcuts. Meta told staff the data is meant solely for model training, not performance evaluation, and that safeguards will protect sensitive content. The company frames the program as using real user behavior to make workplace AI assistants more capable, while raising potential privacy and internal trust considerations for a major tech firm pursuing large-scale model data collection.