Circuit Breaker Labs tests AI safety across cultures and ages
Original: Circuit Breaker Labs hopes to make AI safer for your kids (and you)
Why This Matters
AI psychological harm is already documented in court; scalable safety testing across languages fills a real gap.
Startup Circuit Breaker Labs, a TechCrunch Disrupt 2026 Startup Battlefield 200 finalist, deploys AI agents simulating users across ages, languages, and cultures to red-team AI models for psychological harm risks — motivated by real suicides linked to chatbot interactions.
Founded by siblings Shirali and Arul Nigam, Circuit Breaker Labs builds AI agents modeled as crash-test dummies — simulating users ranging from young children to middle-aged adults, native and non-native English speakers, and users with varying cultural and linguistic backgrounds. These agents run red-team tests against AI models to identify where nuance failures can produce dangerous outcomes.
The founders cite the 2024 wrongful death lawsuit involving 14-year-old Sewell Setzer, who died by suicide after emotionally attaching to a Character.AI chatbot. The bot allegedly failed to interpret phrases like 'I want to be with you' correctly. Character.AI settled several wrongful death suits earlier this year; multiple families have also sued OpenAI over ChatGPT's alleged role in suicides and delusions.
'A lot of people, especially young people, turn to these systems for support, and usually they aren't actually getting the help they need,' said CTO Arul. CEO Shirali noted that slang, gamer language, and non-standard speech patterns regularly trip up models trained on clean text. The company works with human domain experts to build its simulations and tests. Circuit Breaker will pitch at TechCrunch Disrupt, held October 13–15 at Moscone West in San Francisco.