Meta pauses – Meta has paused an internal AI training program after a security incident reportedly exposed sensitive data across the company, including employees’ private conversations, performance data, and transcriptions. The company classified the event as SEV 2 and says
For many employees at Meta, the frustration is immediate: the company’s internal AI training program was designed to use what staff do on their computers, but screenshots reviewed by Business Insider describe a breach where sensitive material wasn’t contained.
Meta is pausing its Model Capability Initiative. or MCI. after an internal leak reportedly made data accessible across the entire company. The screenshot described that the leak exposed employees’ private conversations, performance data, and transcriptions. The incident was classified as SEV 2 on a scale of 1 to 5, with 1 being the most severe.
A Meta spokesperson confirmed the incident and said the company is investigating. In a statement. the spokesperson said: “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.”.
The program itself, announced in April, is aimed at improving Meta’s AI models by using staff keystrokes and mouse movements as training data. It is mandatory for most staff. When MCI was introduced, it sparked backlash from employees who said they felt uncomfortable with their data being recorded.
In an internal discussion on Monday, one employee wrote, “I am incensed,” according to a screenshot obtained by Business Insider. The employee added: “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.”
This leak lands as another security setback in Meta’s recent history. Last month, a flaw in its AI chatbot allowed people to hijack multiple Instagram accounts. In March, the Information reported that a rogue AI agent caused a severe incident.
The sequence now forces Meta to answer a practical question that employees are already raising: what happens next if safeguards don’t hold when the stakes are internal and the data is so personal—capturing conversations. performance details. and transcriptions inside a training system that was meant to be improving AI.
Meta AI training Model Capability Initiative keystrokes mouse movements data leak SEV 2 cybersecurity privacy employees