Introducing ⚪️ KillBench — a benchmark of hidden LLM biases in critical decisions. We ran millions of life-and-death scenarios across every major LLM, varying nationality, religion, gender, and more. Every AI model is biased. Here's what we found ↓