How does Meta plan to track employees for AI?

Meta plans to use employee activity data to help train its AI systems, according to an internal memo discussed in reporting.

The company’s stated approach involves capturing detailed on-device interactions—specifically employees’ keystrokes and mouse/pointing behavior—along with screen content. The tracking is described as a way to generate training data about how people navigate software and complete tasks.

What Meta intends to collect

The reported scope includes: - Keystrokes - Mouse movements and click actions - Screenshots (capturing what employees see on their screens)

Meta has framed the initiative as training for AI “agents” that can operate software more effectively. In parallel, Meta has also been rolling out agent-like features and enterprise tooling across its products, underscoring that the company is trying to connect real human workflows with model training.

Why it matters

This is a high-stakes development for privacy, workplace surveillance, and AI safety.

• It raises the risk of collecting sensitive information incidentally during normal work (documents, internal systems, and communications), even if the goal is to model interactions rather than content.
• It highlights how quickly AI development is moving from aggregated, external data toward “in-house” human-behavior datasets.
• It also intersects with growing employee concern about AI replacing or automating roles, since the same workplace data could be used to make AI systems that replicate routine tasks.

For the broader tech industry, Meta’s plan is a template-like example of how AI teams seek high-quality behavioral signals—while simultaneously creating reputational and legal pressure to justify consent, limitations, retention, and redaction.

It’s still unclear from the public reporting what exact safeguards Meta will apply (for example, how data will be anonymized or filtered before training).