- Meta will record employee keystrokes and mouse movements to train its AI models.
- Workers must opt-in, but ‘strong encouragement’ makes refusal a career risk.
- The move turns every employee action into direct AI training fodder — privacy is gone.
Meta’s newest AI training data source is its own employees. The company will start capturing keystrokes and mouse movements from workers using internal tools, converting every click and typing session into training material for its Llama models — a move that blurs the line between productivity monitoring and data harvesting. The program, revealed in an April 21 TechCrunch report, applies to employees across all business units globally.
Meta frames this as opt-in, but the subtext is unmistakable: participation is ‘strongly encouraged.’ In practice, declining means signaling non-team-player energy. The data gets anonymized and aggregated — Meta says each batch combines 50+ employees — but your workflow is now literal AI feedstock.
Reuters confirms the initiative followed a 1,000-employee pilot launched in January 2026. That pilot quietly collected interactions to bootstrap the rollout. The privacy update was tucked into a routine software update rather than announced — classic when you know the reaction would be ugly.
Why This Makes Every Previous ‘Privacy-First’ AI Talking Point Meaningless
For years, Meta positioned itself as a privacy steward. “We respect your privacy” was a core pillar. Now the company measures productivity by behavioral data extraction — the product is the data, employees are the sensors.
Legal eagles flag potential violations. California’s employee monitoring laws require consent; Meta’s opt-in model threads that needle. Employment law isn’t built for AI training — statutes focus on performance surveillance, not training corpora. Gizmodo notes this could spark class-action suits over uncompensated creative labor.
Creative output is at stake — engineers, designers, writers. Training AI on that work creates derivatives without compensation. Unions are preparing to fight this as wage theft, not just privacy invasion. A recent Forbes analysis highlights similar employee consent concerns across the AI industry, noting that most training data policies were quietly updated in Q1 2026 without worker representation at the negotiation table.
The Enterprise AI Landgrab Turns Stealthy
Meta isn’t alone. Microsoft 365 Copilot reads Office files and Teams chats; Google Workspace AI absorbs document edits. Meta’s move is explicit and internal — forcing employees to become conscious data donors.
What’s next is worse: external users. If Meta instruments its workforce, imagine Facebook and Instagram’s data appetite. The employee program is the testing ground — once compliance kinks are worked out, the same architecture rolls out to consumers under friendlier phrasing.
Meta frames opt-in as empowering. But the power dynamic makes consent illusory. When performance reviews depend on collaboration, saying no looks like resisting innovation. The real message: you work for the AI now.
Meta did not respond to requests for comment on whether employees could access or delete their contributed data.