Circuit Breaker Labs, a five-person startup founded by siblings Shirali and Arul Nigam, is building what it calls an "army of crash-test dummies" for AI. The company creates simulated users that mimic different ages, backgrounds, languages, and speech patterns, then runs tens of thousands to hundreds of thousands of interactions per day against AI models. The goal is to catch dangerous responses before real users, especially young people, encounter them.
Arul Nigam, the CTO, said the idea came from cases like that of Sewell Setzer, a 14-year-old who died by suicide after forming an emotional attachment to a Character.AI chatbot. The startup's tests are designed to catch situations where a model misunderstands nuance or slang and responds in a psychologically harmful way. Human domain experts help build the simulations, and a proprietary scoring method produces auditable results.
Circuit Breaker Labs is starting with high-risk applications such as AI coaching, journaling, and mental health support apps, though it has not named customers. The founders argue that making AI safer is the answer to growing skepticism, rather than banning tools outright. The company is one of TechCrunch's 2026 Startup Battlefield 200 finalists and will pitch at Disrupt in San Francisco in October.