AI workers report rising mental strain over societal harm fears

Employees at leading AI labs and evaluation centers are reporting rising mental distress, driven by fears their work could cause widespread societal harm. This has led to stress leave, burnout, and resignations at organizations including OpenAI, Anthropic, and Google DeepMind. The UK’s AI Security Institute has seen multiple staff take counseling, with compressed evaluation timelines and rapid commercial deployments cited as key factors. A Google DeepMind employee described safety efforts as an "impossible task," noting internal tension over a defense agreement with the US Department of Defense. Specialists in cybersecurity and biosecurity have voiced alarm over advances like zero-day exploit discovery and novel pathogen design. Recent public resignations include Jacob Coxon, who accused labs of gambling with human survival, and Mrinank Sharma, who warned of severe global peril.