- Meta suspended an AI project tracking employee keystrokes after a data leak occurred
- Sensitive employee data was accessible company-wide despite privacy safeguards
- The leak exposed private conversations, performance data, and transcriptions
Meta has suspended an internal AI training project that tracked employee keystrokes, clicks, and mouse movements after sensitive data became accessible company-wide. The leak comes despite Meta claiming at the time of deployment that employee data will not be used for any other purpose and that the tool has 'safeguards in place to protect sensitive content'.
A screenshot of the leak, as accessed by Business Insider, exposed employees' private conversations, performance data and transcriptions. The incident was classified as a SEV 2 on a scale of 0 to 5, with 0 being the most severe. Meta has confirmed the incident, adding that the company was investigating.
"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," a spokesperson told the outlet.
Employees were outraged over the leak, highlighting that Meta did not lock down the data when it was one of the main promises in their original pitch.
"I am incensed," said one employee about the leak in an internal group, as per the screenshot. "I don't see any evidence of malicious access, but the fact that this data wasn't locked down as originally promised is super frustrating."
Humans Training AI Models
In April, the AI training programme called Model Capability Initiative (MCI), developed by the Meta Superintelligence Labs team, was deployed on the workstations of Meta employees. The purpose of the tool, as per a memo to the staff, was to improve the company's AI models in areas where they struggle to replicate how humans interact with computers, like choosing from dropdown menus and using keyboard shortcuts.
"If we're building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them," a Meta spokesperson said at the time.
The programme, which is mandatory for most staff, ignited severe backlash from employees who felt uneasy with their data being recorded.
In a separate memo, Meta CTO Andrew Bosworth said the company would step up internal data collection as part of "AI for Work" efforts, now re-branded as Agent Transformation Accelerator (ATA).
