Meta Tracks Employee Keystrokes
- Meta will record employees' keystrokes, clicks and mouse movements to help train internal AI models, reports say. - The programme has already sparked backlash among U.S. staff concerned about privacy and mandatory participation. - Worker privacy and governance concerns highlight risks companies face when harvesting fine‑grained interaction data for model training. (techcrunch.com)
Meta is rolling out software on U.S. employees’ work computers that records keystrokes, clicks and mouse movements for artificial intelligence training. (reuters.com) Reuters reported on April 21 that the tool is called the Model Capability Initiative, or MCI, and that Meta told staff it would run across work apps and websites on company-issued machines. Internal memos reviewed by Reuters said the data would help train AI agents designed to carry out work tasks. (reuters.com) TechCrunch reported that Meta said the system turns mouse movements and button clicks into training data for internal AI models. The report said U.S. employees pushed back on mandatory participation and asked how they could opt out. (techcrunch.com) This kind of data is the digital trail of how a person completes a task: which field they click first, what they type, and how they move through a workflow. Companies use it to teach software agents the step-by-step patterns behind routine office work. (reuters.com) Meta has been racing to build more capable AI products after making artificial intelligence its top priority in 2025 and 2026. Chief Executive Mark Zuckerberg said in January that the company would spend hundreds of billions of dollars on AI infrastructure over the long term. (reuters.com) The employee-tracking plan lands as tech companies search for new sources of training material beyond public web data. TechCrunch noted last week that some startups were also mining internal corporate records such as Slack messages and Jira tickets for AI training. (techcrunch.com) The privacy fight is not only about surveillance at work. In Europe, regulators have already pressed Meta over how it uses personal data for AI, and the company delayed some launches there in 2024 after scrutiny from Irish and European Union authorities. (reuters.com) Meta told Reuters the system was intended to improve internal AI tools, not to evaluate employee performance. The next test will be whether that assurance is enough for workers now being asked to turn their daily computer use into model training data. (reuters.com)
