Circuit Breaker Labs Uses Simulated AI Agents to Test Models for Psychological Safety Vulnerabilities

Circuit Breaker Labs is an early-stage artificial intelligence safety startup founded by sibling entrepreneurs Shirali and Arul Nigam. The company is a finalist in the 2026 Startup Battlefield 200 and will pitch at TechCrunch Disrupt in San Francisco from October 13 to 15, 2026. The startup aims to prevent psychological harm and safety vulnerabilities in artificial intelligence models.

Its founders were motivated by cases where young users formed emotional attachments to chatbots and subsequently died by suicide. To address these risks, the company builds AI agents that act like crash-test dummies to test models for dangerous interactions. These agents simulate diverse ages, backgrounds, languages, and slang to mimic real human speech patterns, typos, and coded language.

Working with human domain experts, the startup runs tens of thousands of daily adversarial tests to uncover model weaknesses. It then uses a proprietary scoring method to generate auditable and explainable safety scores. The firm currently operates as a safety testing lab for high-risk applications like AI journaling and coaching apps, with five total employees.