Meta is tracking employee keystrokes, mouse clicks, and screenshots on work computers as part of an initiative to train its artificial intelligence models. The program, called the Model Capability Initiative (MCI), collects data from employees interacting with hundreds of websites and apps, including Google, LinkedIn, Wikipedia, Microsoft's GitHub, Salesforce's Slack, and Atlassian. Meta properties like Threads and Manus are also included in the tracking list, which is still evolving.
The data gathered through MCI will be used to train AI models to perform tasks typically done by white-collar workers, such as navigating drop-down menus and using keyboard shortcuts. Meta confirmed the project but did not comment on the specific sites being tracked. The initiative is part of CEO Mark Zuckerberg's push to advance Meta's AI capabilities, following the hiring of Scale AI's Alexandr Wang to lead the effort. Earlier this month, Meta unveiled its first major AI model since Wang's hiring, called Muse Spark, developed by the Meta Superintelligence Labs (MSL).
Employees have expressed concerns about the surveillance aspect of the program, with one worker describing it as 'very dystopian.' Meta has emphasized that the data collection is intended to improve AI models' ability to assist with everyday tasks. CTO Andrew Bosworth has stated that the company's vision includes AI agents handling most work tasks, with humans primarily directing and reviewing their performance.