Workers Hit $10B AI Startup With Data Privacy Lawsuits

A fast-growing AI contractor says it's fueling the industry; a string of workers now accuse it of fueling a data-privacy mess. Mercor, a San Francisco startup valued at around $10 billion that supplies human feedback and other training help to AI giants including OpenAI, Anthropic, and Meta, is facing at least seven recent class-action lawsuits after a third-party data breach allegedly exposed everything from recorded job interviews and facial biometrics to screenshots of contractors' computer screens, the Wall Street Journal reports.

A suit filed in Northern California claims Mercor, which hired 30,000 contractors last year, improperly stockpiled applicant data—like background checks—and shared it with partners. Others say the company monitored workers' devices via screenshot software, reused interview recordings to train AI systems, and encouraged them to share "real-world" examples that may have included proprietary corporate data. Mercor denies wrongdoing, calling the claims "speculative" and saying it follows privacy laws and limits use of confidential information, the Journal reports.

Meta has paused work with Mercor, and sources tell TechCrunch that other large AI firms are reconsidering their relationship with the company. The cases spotlight how AI firms, which have largely exhausted readily available data sources, now hunt for highly specialized, often sensitive data—and how responsibility for where that data comes from can blur as companies race to train models first and sort out rules later. Earlier this month, the Journal reported that Mercor has been offering to pay professionals in multiple industries for their old work materials, despite the fact that such materials generally belong to their employers. (Failed startups have been selling old Slack chats and work emails to AI companies.)