Meta tells U.S. employees it will log computer activity to train AI models
The Facts
- Meta is installing tracking software on U.S.-based employees' work computers.
- The software is intended to capture mouse movements, clicks and keystrokes for use in training Meta's AI models.
- The tool is called Model Capability Initiative, or MCI.
- The software will run on work-related applications and websites and will take occasional snapshots of screen content for context.
- Meta has described the initiative as part of a broader push to build AI agents that can perform work tasks autonomously or help people complete everyday computer-based tasks.
- Meta says the collected data is being used for AI training rather than other purposes, and the company says safeguards exist to protect sensitive content.
- The rollout has prompted internal concern or backlash from some employees, according to reports citing employee comments and internal communications.
Context
What exactly will Meta's new tool collect?
Reports say the software will log employees' mouse movements, clicks and keystrokes on work-related apps and websites, and it will also take occasional snapshots of screen content for context RTE.ie,Reuters.
Why does Meta say it is doing this?
Meta says it needs real examples of how people use computers so its AI agents can better perform everyday computer-based tasks, including actions such as using keyboard shortcuts or navigating menus BBC,Business Insider.
How have employees responded?
The BBC and Business Insider both reported unease among some employees, including objections in internal comments and concerns about helping train systems that could affect future jobs BBC,Business Insider.
How left and right read it
Meta is asking employees to absorb a significant new layer of workplace surveillance in service of its AI ambitions. Recording keystrokes, clicks, mouse movements and occasional screen snapshots across work apps and websites may help build more capable AI agents, but it also raises understandable concerns about consent, boundaries and whether company safeguards are enough for sensitive content.
Meta’s Model Capability Initiative shows how aggressively major companies are moving to build AI tools that can handle routine computer tasks. Using activity from employees’ work machines for training may be a practical way to improve those systems, but the internal backlash underscores a basic management challenge: even when a company says the data is limited to work use and protected by safeguards, trust still has to be earned.
The receipts — all 100 sources
Wire services (6)
Independent coverage (50)
Facts first. Then every angle.
The day’s biggest stories in one short brief — the facts everyone agrees on, then the competing values behind the headlines. Free in your inbox.