Meta to track employee keystrokes for AI training

Meta plans to use data from employee mouse movements and keystrokes to train its artificial intelligence models. The company is launching an internal tool to capture these inputs on certain applications for AI training. The story, which was first reported by Reuters, highlights the lengths to which technology companies are going to find new sources of training data—the information used to teach AI models.

When reached for comment, a Meta spokesperson provided details on the initiative, explaining that the company is building agents designed to help people complete everyday tasks using computers. To achieve this, the spokesperson stated that the models require real examples of how people actually use computers, including mouse movements, clicking buttons, and navigating dropdown menus.

The spokesperson said: “If we’re building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them — things like mouse movements, clicking buttons, and navigating dropdown menus. To help, we’re launching an internal tool that will capture these kinds of inputs on certain applications to help us train our models. There are safeguards in place to protect sensitive content, and the data is not used for any other purpose.”

This trend reveals a “troublesome privacy dimension” of the AI industry. Last week, it was reported that old startups are being scavenged for their corporate communications to be converted into AI training data. This data scavenging includes extracting communications from platforms such as Slack, a corporate communication software, and Jira, a project management and issue tracking software.

Why it matters

This move highlights the aggressive search for new AI training data sources while raising significant privacy questions about the use of internal corporate data for model development.