Companies Just Learned a Brutal Lesson About Training AI to Do Human Jobs

AI staffing startup Mercor said it was hacked through a LiteLLM-related flaw, potentially exposing Slack data and recordings of contractor conversations with AI systems.

The breach raised customer concerns, prompted Meta to pause work, and added fresh pressure on Mercor amid multiple contractor lawsuits.

The incident highlights the security, privacy, and competitive risks in the contractor-driven AI training supply chain used by major AI companies like OpenAI and Anthropic.