Meta to Surveil Staff to Teach Its AI to Work, Report Says
Meta will implement a system to monitor its employees' computer activities, including mouse movements, clicks, keystrokes, and screen content, to train its artificial intelligence systems for workplace applications, according to a new report based on internal company memos.
Tracking Software Details
The initiative, known as the Model Capability Initiative (MCI), will operate on work-related apps and websites, capturing detailed data on how employees interact with digital interfaces. This includes logging mouse movements, clicks, and keystrokes, as well as taking occasional snapshots of what is displayed on employees' screens. The data collection aims to improve AI models in areas where they currently struggle to replicate human-computer interactions, such as selecting items from dropdown menus and utilizing keyboard shortcuts.
An internal memo, posted by a staff AI research scientist in a channel for Meta's model-building SuperIntelligence Labs team, stated that the purpose is to leverage daily work activities to enhance AI performance. "This is where all Meta employees can help our models get better simply by doing their daily work," the memo explained.
AI Integration and Workforce Reshaping
Meta, the parent company of Facebook and Instagram, has been aggressively integrating AI into its workflows as part of a broader strategy to reshape its workforce around the technology. The company argues that this move will increase operational efficiency. In a separate memo shared on Monday, Meta's Chief Technology Officer, Andrew Bosworth, announced an escalation in internal data collection efforts under the rebranded "Agent Transformation Accelerator" (ATA), previously known as "AI for Work."
Bosworth outlined the vision for AI agents to primarily handle work tasks, with human roles shifting to directing, reviewing, and improving these systems. "The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve," he said. He added that the goal is for agents to "automatically see where we felt the need to intervene so they can be better next time." While Bosworth did not specify the training methods, he emphasized Meta's commitment to being rigorous in data collection and evaluation for all types of workplace interactions.
Data Usage and Safeguards
Meta spokesperson Andy Stone confirmed that MCI data would be used as input for AI training but assured that it would not be utilized for performance assessments or other purposes beyond model development. Stone highlighted that safeguards are in place to protect sensitive content, though he did not elaborate on which data types would be excluded from collection.
"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," Stone stated. This approach underscores Meta's focus on leveraging authentic human behavior data to refine AI capabilities in practical work environments.
The report, which includes additional information from Reuters, highlights Meta's ongoing efforts to advance AI technology through extensive internal monitoring, raising questions about privacy and ethical considerations in workplace surveillance.
