Meta halted an internal digital monitoring program after a security breach exposed employees' sensitive data. The system, launched in April, tracked every movement, mouse click, keyboard stroke, and random screenshots of Meta employees and contractors in the United States. Unlike standard productivity tools, Meta aimed to use this human behavior data to train AI models, with CEO Mark Zuckerberg arguing models learn best when they "watch very smart people work." The effort became a "SEV 2 security incident," the company's second-highest severity level, after a permissions failure in Meta's database systems, which include about 45,000 data tables, allowed employees free access to colleagues' keystroke and click data. Exposed information included transcripts, private conversations, employee evaluation data, and sensitive medical and financial information. Meta stated there was no indication the information was exposed externally or misused internally. Before the shutdown, more than 1,600 employees had signed an internal petition warning of security and privacy risks. Management had offered compromises, including the option to pause monitoring for 30-minute windows, but these did not prevent the mass internal data exposure.