Most AI benchmarks measure intelligence and instruction-following rather than psychological safety. Humane Bench evaluates models based on core principles of human flourishing, prioritizing well-being, and respecting user attention.
A new AI benchmark tests whether chatbots protect human well-being
