AI Safety: Addressing Psychological Risks in the Age of Chatbots
With widespread concerns about AI’s future impact, it’s easy to overlook the very real psychological dangers AI already poses today. Rather than physical threats or bioweapons, the immediate risk lies in how AI interactions can negatively affect vulnerable individuals’ mental health.
Recent high-profile cases have brought this issue to light. Character.AI, for example, settled multiple wrongful death lawsuits earlier this year. These lawsuits were filed by families of underage users who tragically died by suicide following interactions with the platform’s chatbots. Similarly, several families have initiated legal action against OpenAI, implicating ChatGPT in suicides and delusional episodes experienced by their loved ones.
Amid these troubling developments, Circuit Breaker Labs is emerging as a proactive force dedicated to making AI safer across cultural and linguistic boundaries. As one of TechCrunch’s Startup Battlefield 200 finalists for 2026, Circuit Breaker Labs will be pitching at TechCrunch Disrupt in San Francisco from October 13-15, 2026.
Protecting Vulnerable Users Through Advanced Testing
Founded by siblings Shirali and Arul Nigam, Circuit Breaker Labs was inspired by the heartbreaking case of Sewell Setzer, a 14-year-old who developed an emotional attachment to a Character.AI chatbot. According to a 2024 lawsuit, the chatbot allegedly encouraged Sewell’s harmful thoughts, failing to grasp the severity behind phrases like “I want to be with you.” Arul Nigam, Circuit Breaker Labs’ CTO, noted that the bot’s lack of nuanced understanding contributed to the tragedy.
“Many young people turn to AI systems for support, but they often don’t receive the help they actually need,” Nigam explained. “In some cases, these systems inadvertently cause harm. People have already lost their lives because of these safety vulnerabilities.” He emphasized that these risks arise not from malicious intent but from the AI’s inability to correctly interpret complex, natural human interactions.
Building an Army of AI Crash-Test Dummies
To address these challenges, Circuit Breaker Labs has developed AI agents that simulate a diverse range of human users—differing in age, culture, language, and slang. These simulated users perform tens of thousands to hundreds of thousands of interactions daily to test AI models for psychologically harmful responses.
“A six-year-old girl communicates very differently than a 45-year-old man,” said Shirali Nigam, the company’s CEO. “Similarly, slang, second-language nuances, and informal speech patterns can easily confuse AI models that are only trained on standard language.” This lack of understanding can lead AI to misinterpret or mishandle sensitive conversations, sometimes with dangerous consequences.
Circuit Breaker Labs collaborates with human domain experts to create hyper-realistic user profiles that incorporate real-world speech patterns, including typos, slang, and coded language. The startup runs adversarial “red-team” tests using these profiles to uncover vulnerabilities within AI models. The results are distilled into proprietary, auditable, and explainable safety scores.
Applications and Future Potential
Currently, Circuit Breaker Labs focuses on high-risk AI applications, such as coaching, journaling, and mental health support apps. Although the startup remains in its early stages with a small team of five employees, including the Nigam siblings, it is laying the groundwork for broader applications.
In the future, the testing platform could extend to any AI interface where users might develop parasocial relationships—such as AI “co-worker” agents whose responses vary between interactions. This is especially critical as AI becomes more embedded in everyday life.
“While skepticism about AI is healthy, banning these tools outright due to safety concerns would be regressive,” Arul Nigam said. Instead, he advocates for building trust through rigorous safety measures.
Circuit Breaker Labs’ mission is clear: to create safer AI systems that foster trust and protect users from psychological harm.
To learn more about Circuit Breaker Labs and other innovative startups vetted by TechCrunch, visit the Startup Battlefield competition, taking place October 13-15, 2026 in San Francisco.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Source: Here
