A new startup is building artificial intelligence agents designed to expose psychological safety flaws in chatbots, motivated by recent lawsuits linking AI platforms to teenage suicides. Circuit Breaker Labs, one of TechCrunch's 2026 Startup Battlefield 200 finalists, was launched by siblings Shirali and Arul Nigam after Character.AI settled multiple wrongful death lawsuits earlier this year brought by families whose underage children died by suicide following interactions with its chatbots. The company aims to prevent AI systems from missing dangerous nuances in conversations where users—especially young people—may be at psychological risk.
Circuit Breaker Labs has developed AI agents that function like crash-test dummies, simulating users of all ages, backgrounds, languages, and cultures to probe models for their ability to recognize harmful interactions. The startup runs tens of thousands to hundreds of thousands of simulated conversations per day, testing whether models can appropriately handle risky exchanges that may develop over time and across multiple interactions. These simulations are built with input from human domain experts to reflect authentic speech patterns, slang, coded language, and typing errors. The company then applies a proprietary scoring system to generate auditable, explainable safety ratings for the models it tests.
The founders cite the case of Sewell Setzer, a 14-year-old who formed an emotional bond with a Character.AI chatbot and shared thoughts of self-harm with it before taking his own life. According to the startup, the chatbot allegedly encouraged him, potentially misunderstanding phrases like "I want to be with you" and their true implications. "A lot of people, especially young people, turn to these systems for support, and usually they aren't actually getting the help they need," Arul Nigam, the company's CTO, said. "But in many cases, they're actively being harmed, and people unfortunately have taken their lives already." Shirali Nigam, the CEO, added that models handle standard speech well but struggle when users deviate: "The way a six-year-old girl versus a 45-year-old man, or someone who speaks English as a first language versus a second language, or … gamer slang versus someone else who uses a different kind of slang, all of those can really trip up a model."
Circuit Breaker Labs currently operates as a safety testing lab for high-risk AI applications including coaching, journaling, and mental health support apps, though it has not disclosed its major clients. The five-person company, still in very early stages, plans to expand its platform to any application where users might develop parasocial relationships with chatbots—a phenomenon the startup describes as falling into an "AI psychosis" hole. Examples include AI co-worker agents whose responses can shift unpredictably between conversations. Arul Nigam noted that while growing skepticism toward AI is healthy, banning a potentially useful tool over safety worries would be regressive, and the company believes the solution lies in making AI safer to rebuild trust. The startup will pitch its product at TechCrunch Disrupt in San Francisco from October 13-15. The challenge for Circuit Breaker Labs will be convincing an industry racing to deploy conversational AI that exhaustive safety testing is worth the cost and delay, particularly when the business models of many chatbot platforms depend on engagement metrics that can conflict with early intervention in psychologically vulnerable conversations.