Meta Pauses Employee Monitoring Tool After Privacy Backlash

Meta has suspended a controversial employee monitoring program that tracked workers' computer activity to gather training data for artificial intelligence models, following internal protests and security concerns.

The program, called the Model Capability Initiative, captured keystrokes, mouse clicks, and on-screen content from staff computers. More than 1,600 Meta employees signed a petition demanding the company halt what they called the harvesting of "employee 'computer use' data," according to The Guardian, which first reported the suspension.

Security exposure amplified concerns

The backlash intensified after Wired reported that data collected through the initiative had been accessible to any Meta employee. An internal security notice referenced exposed data tables containing "full prompts and transcriptions, private conversations, people and performance data."

Meta confirmed the program's suspension in a statement: "We have carefully designed this program with privacy safeguards and while we have no indication at this time that any data was improperly accessed by Meta employees, we're pausing it while we investigate."

Zuckerberg's rationale for employee data collection

At an internal meeting, CEO Mark Zuckerberg explained the program's premise: AI models improve by "watching really smart people do things." He argued that Meta employees possess intelligence and coding skills significantly above average, making their work patterns valuable training material for improving the company's AI capabilities.

The initiative reflects Meta's aggressive AI investment strategy. The company plans to spend up to $145 billion in capital expenditure this year, with substantial portions directed toward AI infrastructure including datacentres.

Why it matters

This incident highlights a growing tension in the AI era: companies seeking proprietary training data are turning inward to their own workforces, creating novel privacy questions that existing workplace policies may not address. The employee petition's emphasis on "consent and trust" suggests that even at technology companies, staff expect boundaries around how their work activity becomes AI training material. The broad internal accessibility of collected data compounds these concerns, demonstrating how monitoring systems can create security vulnerabilities alongside privacy issues.

Separate prediction market project in development

Separately, The New York Times reported that Zuckerberg has directed a small team to develop a prediction market application similar to Polymarket and Kalshi, where users wager on event outcomes. The tentative project, called Arena, would operate independently from Meta's social platforms and remains in early development with no confirmed launch plans.

Mike Proulx, research director at Forrester, characterized the move into prediction markets—an area facing legal scrutiny in the United States—as "not a great look" for a company already under regulatory pressure over its social media products.

The developments were first reported by The Guardian and The New York Times.