Meta $META is pulling back on its program to log employee mouse movements and keystrokes for AI training, after weeks of internal opposition from staff, according to Reuters.
Stephane Kasriel, a vice president in Meta's Superintelligence Labs unit, circulated an internal memo laying out the changes, which include a new option letting employees halt data collection in 30-minute intervals. Full opt-outs will be available to a narrower group, Engadget reports — specifically, remote workers facing bandwidth limitations, employees who regularly handle sensitive information, and those whose work settings make it difficult to keep a laptop on a charger. Battery performance was also addressed, with the team rolling out a series of technical improvements to the software after staff reported it was taxing their machines.
"While we remain confident in the privacy protections we put in place at launch, which went through several layers of risk review, we have heard your concerns about personal data on work devices, battery life, and wanting more control over when capturing happens," Kasriel said in the memo.
The program, known internally as the Model Capability Initiative, was announced last month. The company's stated rationale was that training AI agents to autonomously handle workplace tasks requires authentic data drawn from the way humans actually operate their computers. Employees were not persuaded: opposition among the workforce crystallized into a formal petition that topped 1,500 signatories, the BBC reported. The mood inside the company was sour: One staffer, speaking anonymously to the BBC, called the situation "very dystopian," while a recently departed employee used the episode to voice broader frustration, saying the tool was "just the latest way they're shoving AI down everyone's throat."
At an all-hands meeting last month, Meta CEO Mark Zuckerberg made the case for the initiative, arguing that exposing AI models to skilled workers in action — "watching really smart people do things" — is how those systems get better. In remarks captured in leaked audio and reported by Engadget, he pushed back on privacy concerns: "None of the data is being used for, like, looking at what people are doing, or surveillance, or performance track[ing], or anything like that. It's purely just, like, we are using this to feed a very large amount of content into the AI model."
The tracking initiative landed against a turbulent backdrop. Meta began notifying roughly 8,000 employees of layoffs in May, representing about 10% of its global workforce, as part of a broad restructuring the company has framed as necessary to fund its AI ambitions. Separately, roughly 7,000 additional workers were moved into newly created AI-focused teams. Meta's projected capital expenditures for 2026 run from $125 billion to $145 billion, more than double its 2025 outlay.