wellbeing evaluations
AssessTechniques
Assessment methods designed to measure how AI responses affect user wellbeing and safety.
Why it's here
Placed in Assess: 1 article(s) of evidence from 1 source(s), led by funding news, with 0 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 7Anthropic News·8/25/2026fundingAnthropic launches $5 million grant program for AI wellbeing evaluations
Anthropic is launching a $5 million grant program to support independent research on how AI systems affect user wellbeing. The program provides funding, model access, and technical support for open-source evaluations and benchmarks, with a focus on clinically informed, multi-turn assessments of both harms and overrefusal/overcompliance.