Meta Platforms is currently facing internal protests from its U.S. employees over the mandatory installation of tracking software on their work computers, designed to collect data for training AI agents [2, 4, 5, 8, 16]. This initiative, known as the Model Capability Initiative (MCI), captures mouse movements, keystrokes, and occasionally screenshots to teach AI models how humans interact with computers during daily tasks [3, 5, 16].
Employees have voiced significant privacy concerns and fear that they are inadvertently contributing to the development of AI systems that could ultimately replace their jobs [3, 4, 8, 16]. The protests, which included distributing anonymous flyers in Meta offices, urged colleagues to sign a petition citing the U.S. National Labor Relations Act, arguing that workers are protected when organizing for improved working conditions [2, 4, 8]. Flyers displayed messages such as: “Don’t want to work at the Employee Data Extraction Factory?” [4, 8].
Meta spokesperson Andy Stone stated that the data gathered through MCI is intended solely for “model training” and “would not be used for performance assessments or any other purpose” [2, 16]. The company maintains that safeguards are in place to protect sensitive content [2, 3, 5, 16]. However, employees have expressed discomfort, particularly given that there is no opt-out option for this tracking on company-issued devices in the U.S. [3, 5]. This internal tracking is an extension of existing monitoring practices, though the direct link to AI training and recent layoffs has intensified concerns [3, 5, 16].
Beyond internal operations, Meta also continues to draw scrutiny for its broader data collection practices to train AI. The company utilizes public posts and photos from Facebook and Instagram users for AI model training, with some reports indicating data collection dating back to 2007 [1, 7, 9, 10]. While users in the European Union and the United Kingdom have been granted an opt-out option due to regulations like GDPR, users in the U.S. and Australia do not have this same ability to prevent their public content from being scraped for AI training [1, 7, 9, 15].
Misinformation regarding Meta’s AI training has also circulated, with false claims suggesting the company scrapes direct messages (DMs) for AI purposes [6]. Meta’s updated privacy policy clarifies that it uses interactions with Meta AI tools and related metadata, not private person-to-person DMs, unless those messages are directly shared with Meta AI [6, 15]. Nevertheless, the company faces ongoing criticism regarding transparency, user control, and the ethical implications of using vast amounts of user data for AI development without explicit, universally available consent [1, 10, 14, 15].