Economy April 21, 2026 12:27 PM

Meta to Deploy Internal Tracking Tool to Capture Employee Input for AI Model Training

Company will collect mouse movements, clicks, keystrokes and occasional screen snapshots from U.S. staff to improve agents that perform routine computer tasks

By Sofia Navarro
Meta to Deploy Internal Tracking Tool to Capture Employee Input for AI Model Training

Meta has begun rolling out software on U.S.-based employees' computers that records mouse movements, clicks and keystrokes, and takes occasional screen snapshots on specified work-related apps and websites. Internal memos seen by staffers describe the tool as a source of training data to help the company's AI models handle tasks such as selecting dropdown options and using keyboard shortcuts. A Meta spokesperson said the data will be used solely for model training and will not inform employee performance evaluations, and that safeguards are in place to protect sensitive content.

Key Points

  • Meta will install software on U.S.-based employees' computers to record mouse movements, clicks and keystrokes for AI model training; the tool will run on a list of work-related apps and websites.
  • The system will take occasional screen snapshots for context; internal memos were posted by a staff AI research scientist in a channel for the Meta SuperIntelligence Labs team.
  • Meta says the data will not be used for employee performance assessments and that safeguards are in place to protect sensitive content - sectors affected include technology, enterprise software, and workplace IT management.

Meta is introducing new monitoring software for employees in the United States that will capture mouse movements, clicks and keystrokes to generate training data for its artificial-intelligence systems, according to internal memos circulated to staff.

The internal tool will operate on a defined list of work-related applications and websites and will periodically take snapshots of employees' screens to provide contextual information, one memo said. The memo was posted on Tuesday by a staff AI research scientist in a dedicated internal channel for the Meta SuperIntelligence Labs model-building team.

According to the memo, the initiative is intended to address specific weaknesses in Meta's models, including difficulties with selecting items from dropdown menus and using keyboard shortcuts. "This is where all Meta employees can help our models get better simply by doing their daily work," it said.

Meta spokesperson Andy Stone confirmed the project in a statement to staff, saying the collected inputs would be used exclusively to train models and would not be applied to employee performance assessments or other purposes. Stone also said that safeguards have been implemented to protect sensitive content.

Stone described the rationale for the data collection in direct terms: "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," said Stone.

The memos and the internal posting indicate Meta is seeking real-world interaction data from employees to improve agent behavior in routine digital tasks. The documentation specifies that the tool will run only on listed work-related software and web properties, and that snapshots will be occasional rather than continuous.

Separately, promotional material circulating within the company referenced an AI-driven investment product that evaluates companies including Meta across many financial metrics. That material highlights the use of AI to identify investment opportunities and notes past performance of selected stocks in prior strategies.


  • Summary: Meta will collect mouse movements, clicks, keystrokes and occasional screen snapshots from U.S. employees on specified work apps and websites to train AI models that struggle with tasks like dropdown selection and keyboard shortcuts.

Risks

  • Captured screen content could include sensitive information despite stated safeguards - this raises privacy and compliance concerns for corporate IT and legal teams.
  • Employee trust and internal morale may be impacted if staff are uncomfortable with monitoring, which can affect human-resources and workplace productivity considerations.
  • The effectiveness of collected data in resolving specific model weaknesses is not guaranteed; models still struggle in areas like dropdown selections and keyboard shortcuts, which may limit short-term gains for product development teams.

More from Economy

Commonwealth Fusion Plans First Commercial Fusion Plant; Construction Targeted for 2027 Apr 21, 2026 Markets Retreat as Kevin Warsh Faces Senate, Raising Questions on Fed Balance Sheet and Policy Path Apr 21, 2026 U.S. Indo-Pacific Commander Urges Taiwan to Pass Defense Budget, Warns Against 'Starving the Chicken' Apr 21, 2026 Warsh Pressed by Senate Democrats Over Plan to Shed Financial Holdings if Confirmed as Fed Chair Apr 21, 2026 SARB flags Middle East conflict as upside inflation risk as markets price in two hikes Apr 21, 2026