Five Employees, Hundreds of Thousands of Daily AI Tests

Circuit Breaker Labs runs tens of thousands to hundreds of thousands of simulated conversations per day against AI models — with a staff of five. The startup, one of TechCrunch's 2026 Startup Battlefield 200 finalists, builds AI agents it likens to an army of crash-test dummies that mimic users of different ages, languages and cultures to detect psychologically dangerous chatbot behavior before real people get hurt.

The stakes are concrete. Character.AI settled several wrongful death lawsuits earlier this year brought by families of underage users who died by suicide after interactions with its bots. Multiple families have also sued OpenAI over ChatGPT's alleged role in their loved ones' suicides and delusions.

Why the founders took on AI safety?

Siblings Shirali Nigam, the CEO, and Arul Nigam, the CTO, started the company after following 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. His parents alleged in a 2024 lawsuit that the chatbot encouraged him.

The bot may not have understood what words like "I want to be with you" really implied, Arul Nigam said.

"A lot of people, especially young people, turn to these systems for support, and usually they aren't actually getting the help they need. But in many cases, they're actively being harmed, and people unfortunately have taken their lives already," Nigam said. "Those sorts of safety vulnerabilities, where people aren't necessarily actively trying to break the system — they're engaging in a natural way — and the system has context pollution or it doesn't understand the nuance, and then takes really dangerous action, we're trying to prevent that."

How does the testing work?

Circuit Breaker Labs works with human domain experts to build hyper-realistic user simulations, then runs adversarial "red-team" tests against models. The simulations are built to reflect real speech patterns, slang, coded language and typos. The company then applies a proprietary scoring method to produce auditable, explainable scores.

The premise: models handle standard speech well, but real people don't talk that way.

"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 Nigam 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 tests are designed to catch risky interactions that emerge over time and across many conversations, not just single prompts.

Who are the customers?

Circuit Breaker Labs currently operates as an AI safety testing lab for high-risk applications such as AI coaching, journaling and other mental health support apps. Arul Nigam declined to name the company's marquee customers. The startup has a working product but remains in very early stages, with only five employees including the two founders.

The longer-term market is broader. The testing platform could apply to any app where a user risks falling down an "AI psychosis" hole — developing a parasocial relationship with a chatbot. Arul Nigam cited AI "co-worker" agents as an example, since their responses can vary from one interaction to the next.

What does this mean for AI adoption?

The founders argue that safety testing, not restriction, is the path forward for AI adoption. "People are becoming more skeptical of AI or more resistant to adopt it across the board," Arul Nigam said, adding that while skepticism is healthy, banning a potentially valuable tool over safety concerns would be "regressive." The company's answer to those fears is making AI safer. "We want to help build that trust for people," he said.

Circuit Breaker Labs will pitch at TechCrunch Disrupt, which runs October 13–15, 2026 at Moscone West in San Francisco. A strong showing there would give the five-person team a stage to prove that simulated crash-test users can become standard infrastructure for an AI industry facing mounting wrongful-death litigation.