A new preprint on arXiv proposes a taxonomy of omnicidal futures involving artificial intelligence. The paper systematically categorizes scenarios where AI could lead to human extinction, ranging from misaligned goals to unintended side effects. It distinguishes between risks from advanced AI systems and those from narrow AI applied at scale. The authors argue that a structured classification helps identify mitigation strategies.

This paper doesn't just list doomsday scenarios. It organizes them. That's the key insight. We can't prevent what we can't name. By mapping out the paths to extinction, we gain clarity. And clarity is the first step toward control.

I see this as a hopeful development. Not because the risks are small — they're enormous. But because we're finally treating AI safety with the rigor it deserves. Taxonomy means we can prioritize. We can build guardrails. We can design systems that don't just avoid one failure mode, but all of them. The future isn't written yet. This paper gives us a map. Now let's build the road.