Now Meta will track what employees do on their computers to train its AI agents

2026-04-22 · Source: The Verge · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, quick

Summary

Meta is deploying a new internal tool called Model Capability Initiative (MCI) on US-based employee computers to collect data for training its AI agents. This tool records mouse movements, clicks, keystrokes, and occasional screenshots from work-related applications and websites. The collected data aims to improve AI models' ability to interact with computers like humans, specifically for automating tasks similar to those performed by Meta employees. A Meta spokesperson confirmed the initiative, stating safeguards are in place for sensitive content and that the data will not be used for performance assessments. Meta CTO Andrew Bosworth outlined a vision where AI agents primarily perform work, with employees directing and refining their performance.

Key takeaway

For CTOs and VPs of Engineering exploring AI-driven automation, Meta's MCI initiative highlights a direct approach to acquiring high-fidelity training data. You should consider internal data collection strategies, ensuring robust privacy safeguards and clear communication, to accelerate your AI agent development and align with a vision of agents performing core operational tasks.

Key insights

Meta is using employee activity data to train AI agents for task automation.

Principles

• Real-world human interaction data is crucial for AI agent training.
• AI agents are envisioned to become primary task performers.

Method

The MCI tool captures mouse movements, clicks, keystrokes, and screenshots from employee computers in work apps to generate training data for AI models.

In practice

• Collect interaction data from internal tools.
• Focus AI training on automating routine tasks.

Topics

• Model Capability Initiative
• AI Agent Training
• Employee Activity Tracking
• Workplace Automation
• Agent Transformation Accelerator

Best for: CTO, VP of Engineering/Data, Executive, AI Engineer, MLOps Engineer, AI Ethicist

Related on AIssential

• Meta tracks US employees' clicks and keystrokes to train AI agents — Meta is tracking employee computer activity to train AI agents for task automation, coinciding with workforce reductions.
• Meta just did the thing — AI agents operating in virtualized environments demonstrate significant potential for automating complex remote work tasks.
• Meta to Train AI Through Tracking Workers’ Movements on PC — Meta is tracking employee computer activity to train AI agents, raising significant privacy and job security concerns.
• the SCARIEST chart in AI — AI agent capabilities are accelerating, replacing human expert labor at an increasingly rapid pace.
• Meta will record employees’ keystrokes and use it to train its AI models — Companies are increasingly using internal employee interaction data to train AI models for improved task execution.
• BREAKING: Meta buys Manus AI — Meta's acquisition of Manis AI highlights the growing importance of agentic systems and the "model capability overhang problem."

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Verge.