Meta Employee Surveillance Row
- Reports say Meta installed mandatory tracking software that captures employee keystrokes and activity to train AI agents. - Staff have protested the surveillance, claiming the tracking is mandatory with no opt-out. - The backlash highlights tension between collecting behavioural training data and workplace privacy, risking reputational and regulatory fallout. (theregister.com)
Meta has started installing software on U.S. employees’ work computers that records keystrokes, mouse movements and clicks to train artificial intelligence systems. (reuters.com) The tool is called Model Capability Initiative, or MCI, and internal memos reviewed by Reuters said it will collect activity across work-related apps and websites. Meta said the project is meant to help build AI agents that can carry out office tasks on a computer. (reuters.com) CNBC reported on April 22 that the monitored list includes hundreds of sites and apps, including Google, LinkedIn and Wikipedia. CNET and PCMag reported that the software can also capture occasional screenshots or “screen content” from employee devices. (cnbc.com) (cnet.com) (pcmag.com) The system is aimed at a specific artificial intelligence problem: teaching software agents how people actually use a computer. Instead of learning only from text, these models learn from sequences of clicks, shortcuts, menus and typed commands. (reuters.com) (computerworld.com) Employees pushed back after learning the program was mandatory and had no opt-out, according to Business Insider and other follow-up reports. One of the recurring objections in internal discussions was that staff did not want detailed behavioral data collected from routine work without a choice. (businessinsider.com) (entrepreneur.com) Meta said the data will not be used for performance reviews, and spokesperson Andy Stone told Reuters that employee activity on company machines has been monitored “in some capacity” for years. The company’s position is that MCI expands existing monitoring for model training rather than worker evaluation. (reuters.com) The dispute lands in the middle of Meta’s broader push to automate work with generative artificial intelligence. Chief Executive Mark Zuckerberg has spent the past year reorganizing the company’s AI efforts and pitching software agents that can handle more tasks across apps and services. (techcrunch.com) (computerworld.com) The fight also puts a workplace issue in front of regulators who already scrutinize Meta on privacy and data handling. A program that turns staff behavior into training data gives the company more realistic examples for its models, but it also creates a fresh record of who did what on a work machine and when. (oecd.ai) (computerworld.com) For now, the immediate question inside Meta is narrower than the company’s AI ambitions: whether workers will have to keep feeding the system every time they type, click or move a mouse on a company laptop. (reuters.com)
