Meta to Harvest Employee Keystrokes and Mouse Movements in Radical AI Workforce Overhaul

Meta begins capturing employee clicks and mouse movements to train AI agents, while 8,000 job cuts loom. See the future of white-collar AI automation.

By: AXL Media

Published: Apr 22, 2026, 8:17 AM EDT

Source: Information for this report was sourced from Reuters

The Architecture of Total Workplace Surveillance

Meta Platforms is deploying a sophisticated tracking tool across its U.S. workforce designed to log every mouse movement, click, and keystroke made by employees on their company devices. According to internal memos from Meta’s SuperIntelligence Labs, this Model Capability Initiative (MCI) will also take periodic snapshots of employee screens to provide a visual context for the captured data. The objective is to provide the company’s "Muse Spark" AI models with high-fidelity training data on how human professionals navigate dropdown menus, use keyboard shortcuts, and manage complex office tasks.

Transforming the Workforce into AI Training Data

This aggressive data collection is a cornerstone of a re-branded internal program called the Agent Transformation Accelerator (ATA), formerly known as "AI for Work." Meta CTO Andrew Bosworth informed staffers that the ultimate vision is a workplace where AI agents perform the primary labor while humans shift into oversight roles. By capturing the minute-by-minute workflows of its own software engineers and analysts, Meta intends to build "AI builders"—autonomous agents capable of building, testing, and shipping future products with minimal human intervention.

Mass Layoffs and the Efficiency Pivot

The technological shift comes at a high human cost, as Meta prepares to eliminate approximately 8,000 roles, or 10% of its global workforce, beginning on May 20, 2026. Unlike the "Year of Efficiency" cuts of previous years, these layoffs are specifically tied to an AI-driven restructuring. Meta’s leadership has set ambitious targets for its remaining staff, with some engineering teams expected to generate 75% of their code using AI tools by mid-2026. CEO Mark Zuckerberg has publicly indicated that these advancements will eventually allow individuals or small teams to perform work that previously required entire departments.