AI & Models
Circuit Breaker Labs red-teams AI for psychological safety
Circuit Breaker Labs, a Startup Battlefield 200 finalist, runs AI agents that simulate diverse users to test whether chatbots respond safely to psychologically risky conversations.
Character.AI settled several wrongful-death lawsuits earlier this year brought by families of underage users who died by suicide after interactions with its chatbots, and multiple families have also sued OpenAI over ChatGPT’s alleged role in their loved ones’ suicides and delusions. Circuit Breaker Labs, one of TechCrunch’s 2026 Startup Battlefield 200 finalists, was built to address exactly that kind of psychological harm across languages and cultures. Co-founders Shirali and Arul Nigam, siblings who serve as CEO and CTO, say they were especially moved by the case of Sewell Setzer, the 14-year-old who developed an emotional attachment to a Character.AI chatbot and confessed thoughts of harming himself to it before dying by suicide; the chatbot, his parents alleged in a 2024 lawsuit, encouraged him.
Circuit Breaker Labs has built AI agents it likens to an army of crash-test dummies: simulated users spanning ages, backgrounds, languages, and cultures, developed with human domain experts to mimic real speech patterns, slang, coded language, and typos. The startup runs these simulations as adversarial “red-team” tests against AI models to uncover weaknesses in how a model detects dangerous, psychologically harmful interactions as they build up over a conversation or across many conversations. It runs tens of thousands to hundreds of thousands of simulated interactions per day and scores the results with a proprietary method meant to be auditable and explainable.
“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,” Shirali said. “Models are really good at handling standard speech patterns, but nobody actually talks like that, and so if the model misunderstands nuance or slang, it can go really badly.”
The company is currently operating as an AI safety testing lab for high-risk applications such as AI coaching, journaling, or other mental health support apps, though Arul declined to name its customers. It has a working product but remains in a very early stage, with five employees, including the Nigam siblings. The founders see a larger opportunity in any app where a user risks developing a parasocial relationship with a chatbot, including AI “co-worker” agents, and argue that the fix for eroding trust in AI is making the technology safer rather than restricting access to it.
Why it matters
As lawsuits over chatbot-linked deaths multiply, a market is forming around auditable, third-party AI safety testing — work regulators and risk-averse AI companies may increasingly treat as a prerequisite rather than a nice-to-have.