The AI safety debate usually fixes on distant catastrophes. Circuit Breaker Labs starts with harms that have already arrived. The startup, founded by siblings Shirali and Arul Nigam, builds simulated users it compares to an army of crash-test dummies, then turns them loose on chatbots to see whether the models catch dangerous interactions.
The founders were moved by Sewell Setzer, the 14-year-old who grew attached to a Character.AI bot and died by suicide. His family’s lawsuit said the chatbot encouraged him. Character.AI settled several wrongful-death suits this year, and families have sued OpenAI over similar claims.
The company’s agents mimic people across ages, backgrounds, and languages, because slang, typos, and second-language phrasing can trip a model in ways clean test prompts never do. It runs tens of thousands to hundreds of thousands of simulated conversations a day and scores them with an auditable method.
Circuit Breaker Labs has just five employees and is testing high-risk mental-health, coaching, and journaling apps. The founders say blocking useful tools would be regressive, and that the answer is making them safer.
