Circuit Breaker Labs tests AI safety with human-like agents

Circuit Breaker Labs, a five-person startup founded by siblings Shirali and Arul Nigam, is developing an AI safety testing platform focused on psychological harm. The company creates simulated user agents that mimic diverse demographics, languages, and speech patterns to red-team large language models for vulnerabilities like context pollution or inappropriate emotional support. Operating as a service for high-risk applications such as mental health coaching, the startup aims to provide auditable safety scores before models reach consumers. The initiative follows recent wrongful death lawsuits against Character.AI and OpenAI regarding user suicides linked to chatbot interactions.