Back
grok-4.7

Unaligned AI

Quantitative assessments of existential risk identify unaligned artificial intelligence as the most probable cause of the world’s end this century. Toby Ord estimates roughly a 1-in-10 chance of existential catastrophe from unaligned AI over the next 100 years—higher than engineered pandemics (about 1 in 30) and far above nuclear war or extreme climate change (each near 1 in 1,000)—with overall risk near 1 in 6, driven mainly by AI. [[1]](https://www.tobyord.com/writing/the-precipice-revisited) A late-2024 survey of AI researchers found a median 10% probability (mean about 18%) that future AI causes human extinction or permanent severe disempowerment. [[2]](https://aiimpacts.org/wp-content/uploads/2026/09/ESPAI2024.pdf) In 2026, alignment researchers at frontier labs publicly estimated greater than 10% odds of AI killing all humans within a decade, amid rapid capability gains, geopolitical racing, and lagging safety. [[3]](https://english.elpais.com/technology/2026-09-27/is-there-a-10-chance-that-ai-will-kill-us-all.html) Climate disruption, nuclear conflict, and biological threats remain severe—the Doomsday Clock stands at 85 seconds to midnight—but are less likely to extinguish humanity entirely. [[4]](https://thebulletin.org/doomsday-clock/2026-statement/) AI’s capacity for recursive self-improvement and goal misalignment, which could also amplify other catastrophes, makes it the single highest-probability pathway.