After over 1,500 Meta employees signed a petition, the social media giant has scaled back its controversial Model Capability Initiative (MCI) — but has not stopped it entirely. The conflict highlights the growing tension between data privacy, AI training, and employee rights inside tech companies.
What is MCI and why did it spark outrage?
Launched in April, MCI software was installed on US-based Meta employees' work laptops. It records mouse movements, clicks, keyboard inputs, and screenshots across hundreds of websites and apps including Google, LinkedIn, GitHub, Slack, and Wikipedia. The data feeds Muse Spark, a cutting-edge multi-step workflow system developed by Meta's Superintelligence Labs, designed to train AI to use computers like a smart human.
CEO Mark Zuckerberg stated at an all-hands meeting that employee-generated data is far higher quality than outsourced labelers. However, after deployment, employees reported sluggish performance and significantly reduced battery life. Some workers posted flyers in conference rooms and vending machines, citing the National Labor Relations Act to demand improved working conditions, labeling Meta an "Employee Data Extraction Factory". Despite Zuckerberg's assurance that MCI data would only be used for AI training and not affect performance reviews, the backlash grew, culminating in a petition signed by over 1,500 employees.
Meta's four concessions
In an internal memo obtained by The Information and confirmed by Reuters, Meta Vice President Stephane Kasriel announced four changes to MCI:
- Employees can pause the tracker for 30 minutes at any time to handle personal matters;
- Data collection will shift from recording specific keystrokes to activity summaries;
- Broader exemption channels for employees handling sensitive content, remote workers with limited bandwidth, and field workers with charging difficulties;
- Multiple optimizations have been implemented to reduce battery drain.
Meta also emphasized that only a "very small number of engineers" have access to raw data. However, the memo made clear that MCI has not been terminated — only modified. The tracking mechanisms remain operational and still cover a wide range of activities.
Regulatory vacuum and layoffs fuel anxiety
Legally, the US has no federal law prohibiting employers from monitoring work computers. Meta deliberately limited MCI deployment to American employees to avoid strict EU GDPR scrutiny on systematic behavioral surveillance. Notably, Meta is simultaneously laying off approximately 8,000 employees — 10% of its global workforce. The combination of massive layoffs and using remaining employees' behavioral data to train AI creates a volatile mix. From a crypto and Web3 perspective, this conflict underscores the battle between centralized data control and personal data sovereignty. Decentralized solutions — such as on-chain identity and local data processing — could offer an alternative where data ownership returns to users rather than being concentrated on corporate servers. For now, Meta's concessions provide temporary relief, but the surveillance infrastructure remains in place. Similar labor struggles will likely become more frequent in the AI era.