The program was launched earlier this year by the company’s AI unit, TBD, which is comprised of approximately 100 researchers. While Meta leadership argued that the data was necessary to improve AI proficiency in basic functions—such as navigating dropdown menus and utilizing keyboard shortcuts—the project triggered immediate resistance from staff members who cited significant privacy and security risks.
Internal Opposition and Security Concerns
The controversy intensified after employees discovered that the program did not initially offer an opt-out mechanism. Concerns grew among the workforce regarding the potential for the collection of sensitive information, including personal passwords, immigration documentation, and financial details. The internal backlash manifested in various forms, with workers circulating flyers, sharing memes, and labeling the program’s screen-recording icon the “Eye of Sauron.”
A petition calling for the program’s termination garnered more than 1,800 signatures. Although Meta attempted to address these concerns in June by allowing employees to pause tracking for 30-minute intervals during sensitive tasks, staff maintained that the measure was insufficient and continued to demand the total cancellation of the initiative.
The situation reached a breaking point in late June when an engineer discovered that the collected data, which reportedly included private conversations, was accessible across the company. This revelation contradicted earlier assurances from Meta leadership, including Chief Technology Officer Andrew Bosworth, who had previously described the data as “tightly controlled.” Following an internal post highlighting the security lapse, the company suspended the program within hours.
Conflicting Perspectives on the Initiative
While Meta has stated that the program is suspended rather than permanently cancelled, the company indicated that any future iteration would be made optional. The extent to which the collected data was utilized remains unclear; while one manager suggested in July that the information had not been used to train the company's latest models, Bosworth reportedly remarked during a town hall that the data had proven more useful than anticipated.
The fallout from the initiative extended to internal management relations. In a leaked audio recording from July, Bosworth criticized the employees who spearheaded the opposition, suggesting that the campaign had hindered the company’s ability to implement necessary changes. Conversely, labor advocates have characterized the successful effort to halt the program as a significant development in workplace privacy, noting that Meta’s attempt to gather such data was ultimately thwarted by the collective action of its staff.
