Meta ended a program tracking employees' computer movements to train AI models over security concerns, according to a new report.
Wired detailed that an internal security issue exposed potentially sensitive data to company workers as a result of the program.
"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," company spokesperson Tracy Clayton told the outlet.
The program in question tracks employees' mouse movement and keystrokes to train its AI model. Unveiled in late April and called Model Capability Initiative (MCI), it also takes snapshots of employees' screens.
A Staff AI research scientist sent a memo to the rest of the company, saying the goal is improving the company's AI models in areas where they are struggling to replicate the way in which humans interact with computers.
"This is where all Meta employees can help our models get better simply by doing their daily work," reads a passage of the document.
In another memo, Meta CTO Andrew Bosworth said that the company will increase internal data collection efforts. "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.
Meta spokesperson Andy Stone said the model needs "real example" of how people use computers for every day tasks to help develop AI. "Things like mouse movements, clicking buttons, and navigating dropdown menus," he said.
Employees promptly launched a protest against the program. Employees distributed flyers across U.S. offices encouraging colleagues to sign an online petition against the initiative. "Don't want to work at the Employee Data Extraction Factory?" the flyer said, according to a May report.
The petition in question claims that "when employees asked what privacy reviews were conducted, including any 'people data reviews' (which are required for processing employee data), no completed privacy reviews were provided."
"The outlined privacy mitigations were vague, and leadership's confidence in them appeared limited - evidenced by the selective opt-out afforded to executives," it adds.
A former employee who pushed back against the program told Wired that the security lapse was a "mess" and colleagues expected it to take place.
"When workers raised concerns, leadership doubled down and failed to acknowledge the risks workers raised about the safety and privacy of worker and customer data," the person said.
"Leadership has clearly created an authoritarian environment where workers are no longer respected or heard."
Stephane Kasriel, a Meta vice president overseeing AI research, told staff the issue was addressed less than four hours after being found. However, the initial fix did not work. She said the company will resume the program "when we are confident in the effectiveness of our data protection control," but noted that it has already "gathered sufficient data to assess the long-term value of the tool."
