Meta tracks employee input

- Meta will start capturing employees' mouse movements, clicks and keystrokes on U.S. work devices for AI training. - The program is part of Meta's reworked 'Meta Superintelligence Labs' effort and has provoked angry employee reactions. - The rollout raises telemetry, privacy and data-pipeline design questions for training workplace agents (reuters.com).

Meta is starting to record many U.S. employees’ clicks, mouse movements and keystrokes on work devices to build new artificial intelligence systems. (reuters.com) Reuters reported on April 21 that the software will run on work-related apps and websites, and internal memos said it can also take occasional snapshots of what is on employees’ screens. Meta said the data is for training models that can handle work tasks on their own. (reuters.com) The tool is called Model Capability Initiative, or MCI, and the rollout applies to U.S.-based employees’ computers. In a separate memo, Chief Technology Officer Andrew Bosworth said Meta would expand internal data collection under an “AI for Work” program now renamed Agent Transformation Accelerator. (reuters.com) The basic idea is simple: if a model can watch how people complete a task on a computer, it can learn the sequence of steps the way a trainee watches a coworker. Mouse paths, clicks, typed commands and screen changes become a record of how work gets done across software tools. (techcrunch.com) That puts the program in the middle of a larger race to build “agents,” or software that can carry out office tasks with less human input. Reuters said Meta described the effort as part of a broader push to create systems that can perform work autonomously. (reuters.com) The timing is tight because Meta has been rebuilding its artificial intelligence group around Meta Superintelligence Labs. On April 8, the company introduced Muse Spark, the first model from that unit, and said the lab had rebuilt its AI stack over the prior nine months. (about.fb.com) Meta put former Scale AI chief Alexandr Wang in charge of Meta Superintelligence Labs after investing $14.3 billion for a 49% stake in Scale AI, according to TechCrunch’s April 8 report on the reorganization. CNBC reported the same day that Muse Spark was the first major model release under Wang. (techcrunch.com) (cnbc.com) Employees reacted angrily in internal discussions reviewed by Reuters, which said some staffers questioned why the company was collecting such detailed behavior data and whether the systems being trained could eventually replace some of the people generating it. Meta said the program was not being set up for performance reviews. (reuters.com) Those fears land as Meta prepares another round of cuts. Reuters reported on April 17 that the company was targeting May 20 for a first wave of layoffs affecting about 10% of its global workforce, or nearly 8,000 employees, with more cuts later in 2026. (reuters.com) The immediate question is not whether Meta can collect the data on its own machines, but how much of workers’ daily activity it can safely turn into training material without sweeping in passwords, private messages or sensitive business information. The opening move is clear: Meta wants its next workplace agents to learn from the people already doing the jobs. (reuters.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.