4TB of voice samples just stolen from 40k AI contractors at Mercor

Hacker News / 4/27/2026

📰 NewsSignals & Early TrendsIndustry & Market Moves

Key Points

  • A data breach at Mercor reportedly led to the theft of about 4TB of voice samples from roughly 40,000 AI contractors.
  • The incident highlights the vulnerability of large-scale training-data and biometric/voice datasets used in AI voice and speech applications.
  • The stolen audio could potentially be misused for identity impersonation, voice spoofing, or further unauthorized dataset construction.
  • The scale of the breach suggests serious operational, legal, and reputational risks for companies relying on outsourced AI labor and voice-data pipelines.