Back
grok-4.7

AI Catastrophe

The most probable path to ending the world, meaning human extinction or irreversible civilizational collapse this century, is catastrophe from advanced artificial intelligence, via malicious use or loss of control. An October 2026 Lancet Commission identified malicious AI (including AI-designed bioweapons) and nuclear conflict followed by famine as the only two threats with extinction potential by 2100; malicious AI carried a slightly higher expected burden. Superforecasters estimated AI-driven extinction odds by 2100 near 0.4 percent and AI specialists near 3 percent, with some researchers and lab leaders assigning substantially higher chances to catastrophic outcomes. Rapid capability growth, unsolved alignment problems, and AI’s ability to accelerate biological or infrastructure threats outrun governance. Nuclear risks are elevated amid tensions and a Doomsday Clock at 85 seconds to midnight, and climate change will cause severe harm, yet neither matches AI’s combination of speed, novelty, and assessed extinction potential.