322
Audio & Video Production299
Software Development231
Automation & Workflow200
Writing & Content Creation183
Marketing & Growth175
AI Infrastructure & MLOps143
Design & Creative144
Data & Analytics110
Photography & Imaging139
Voice & Speech123
Customer Support113
Sales & Outreach108
Education & Learning116
Operations & Admin79
Meta will capture mouse movements, clicks, and keystrokes from US employees on some work tools to create training data for AI agents, reports say.
In short: Meta plans to use employee tracking software on some work tools in the US to collect data that helps train AI agents.
Reuters reports that Meta will start capturing US employees’ mouse movements, clicks, and keystrokes as they do day to day work. The report cites internal memos describing a “Model Capability Initiative” run by Meta Superintelligence Labs.
According to the memo, the tracking will run on specific work-related apps and websites. It will also take periodic screenshots to add context, like a snapshot of what was on screen at the time.
Meta spokesperson Andy Stone told Reuters the goal is to improve AI agents, meaning AI tools that can use a computer for you (like a helper who can click buttons and fill out forms). Stone said this kind of training data can help with tasks AI sometimes struggles with, such as navigating menus and dropdown lists. He also said the data will not be used to evaluate employees.
The report notes that tracking employees in Europe could conflict with national laws that limit how companies monitor workers. Meta has faced scrutiny in the European Union before over how it handled opt-outs for using people’s content to train AI.
Many new AI agents are being built to “drive” a computer the way a person does. To learn that, they need examples of real clicks and steps, not just text from the internet (like teaching someone to cook by watching what you do in the kitchen, not only reading recipes). Meta’s approach could make these tools more capable, but it also raises privacy and workplace monitoring questions, especially as more companies look for similar data.
Source: Arstechnica