In all the discussions about how AI might one day destroy humanity, it’s easy to forget that AI has already posed a threat to some people’s lives—not through biological weapons, but on a psychological level.

For example, earlier this year, Character.AI reached settlements in several cases of wrongful death lawsuits. These lawsuits were filed by the families of minor users, who claimed that the relevant users committed suicide after interacting with their robots. There are also multiple families suing OpenAI, accusing ChatGPT of playing some role in the suicides and delusions of their relatives.

Making AI safer in different language and cultural environments is the mission of Circuit Breaker Labs. The company is one of the finalists for TechCrunch 2026 Startup Battlefield 200. (Circuit Breaker will participate in the TechCrunch Disrupt roadshow. This event will be held in San Francisco from October 13th to 15th, 2026 Moscone West.)

The founders, Shirali Nigam and Arul Nigam, are siblings. A case that motivated them to start a business was Sewell Setzer – a 14-year-old boy who developed an emotional attachment to a chatbot named Character.ai and confided his thoughts about harming himself to it before committing suicide. According to a lawsuit filed by his parents in 2024, this chatbot also encouraged him. Arul Nigam, the chief technology officer of Circuit Breaker Labs, stated that this chatbot may not truly understand the underlying meaning of statements like “I want to be with you.”

Arul Nigam said, "Many people, especially young people, seek support from these systems, but often they do not receive the help they truly need. However, in many cases, they are actually being harmed, and unfortunately, some have even taken their own lives as a result."

He also said, "These security vulnerabilities are not about people actively trying to break into the system—they are just interacting with it in a natural way—but the system can suffer from context contamination, or it may fail to understand the subtle differences, and then respond in a very dangerous manner. We are trying to prevent this from happening."

Circuit Breaker Labs has developed a AI proxy system, which it likens to an army of "crash test dummies." These proxies simulate people of different ages, backgrounds, languages, and cultures to test the models' ability to recognize dangerous and psychologically harmful interactions.

Circuit Breaker Labs CEO Shirali Nigam said, "Differences in the way a six-year-old girl and a 45-year-old man speak, between those whose English is their native language and those who use it as a second language, or between those who use gaming slang and those who use another type of slang, can all cause the model to make mistakes. The model is very good at handling standardized speech patterns, but in reality, no one speaks in that way. So if the model misunderstands the nuances of tone or slang, the consequences could be serious."

This startup company collaborates with experts in the human domain to create highly realistic user simulations for conducting “red team” testing on models. Red team testing refers to adversarial testing aimed at identifying weaknesses. These tests are designed to mimic real human speech patterns, slang, jargon, and spelling mistakes. Subsequently, Circuit Breaker Labs runs tens of thousands to hundreds of thousands of simulated interactions every day.

The goal is to ensure that the model can respond appropriately to high-risk interactions that may gradually emerge over long periods of time and multiple rounds of dialogue. Afterwards, Circuit Breaker Labs will use a proprietary scoring method to generate auditable and explainable scoring results.

Currently, Circuit Breaker Labs operates as a security testing laboratory for AI, serving high-risk scenarios such as AI coaches, AI diaries, and other mental health support applications. However, Arul Nigam refuses to disclose the names of its iconic clients. Although this startup already has a functional product, it is still in a very early stage; including Nigam and her brother, the company currently has only 5 employees.

However, in the long run, this testing platform can also be applied to any scenario that may lead people into the deep pit of “AI mental illness,” that is, in situations where there is a risk of forming pseudo-social relationships during interactions with chatbots. Examples include AI “colleague” proxies, as their responses may vary with each interaction.

Arul Nigam said, "People are becoming more skeptical about AI or more resistant to its full adoption." He added that skepticism is healthy, but banning a potentially valuable tool solely due to security concerns would be a "regression."

Circuit Breaker Labs believes that the answer to these concerns is to make AI more secure. "We hope to help people build this trust."

According to TechCrunch, if you want to learn more about Circuit Breaker Labs and many other innovative startups that have been screened by TechCrunch, you can pay attention to the Startup Battlefield competition that will be held in downtown San Francisco from October 13th to 15th.