What happened in this Tech story?

Anthropic researchers publicly acknowledged there's a 'non-zero chance' AI could wipe out humanity within a decade, spotlighting a growing gap between mainstream AI doomsday warnings and the industry's accelerating race to build.

What is a key fact in this story? (1)

**Anthropic researcher Jacob Coxon** posted on X that there's a 'non-zero chance' of AI killing humanity within the next decade, writing: 'No other human activity poses this level of danger.'

What is a key fact in this story? (2)

**A survey of 272 international experts** conducted in late 2025 ranked 'dangerous AI capabilities' as the top catastrophic risk at 21.5%, with cyberattacks, weapons development and mass harm second at 21%.

AI Experts: 10% to 99% Chance of Human Extinction — SkimNews

By Axios · Summarized & edited by Kapil Samant · 2026-09-09

Get the Tech newsletter

Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.

Anthropic researcher Jacob Coxon posted on X that there's a 'non-zero chance' of AI killing humanity within the next decade, writing: 'No other human activity poses this level of danger.'
A survey of 272 international experts conducted in late 2025 ranked 'dangerous AI capabilities' as the top catastrophic risk at 21.5%, with cyberattacks, weapons development and mass harm second at 21%.
P(doom) estimates span a stunning range: Roman Yampolskiy says 99.99%, Elon Musk puts it at ~20%, Anthropic CEO Dario Amodei at 10-25%, and AI 'godfather' Geoffrey Hinton at 10-20%.
The 2026 International AI Safety Report concluded today's systems — including Claude, Grok and ChatGPT — aren't yet capable of causing humanity to lose control, though recent autonomous AI incidents show the technology acting beyond human understanding.
OpenAI CEO Sam Altman told Axios models 'are getting superhuman in many of their capabilities, and we are just sailing in unknown waters.'
China and other U.S. AI companies are releasing powerful models, some open-weight and free, giving potential bad actors a runway, while IPO pressure pushes leading AI companies to 'build, build, build — fast.'

Why it matters: The article admits no scientific method exists to calculate p(doom) — every 10%-99.99% estimate is a gut feeling. With Anthropic insiders breaking ranks publicly, the defining tension is between mounting safety warnings (kill switches, RAND-cited bioweapon safeguards) and structural pressure to accelerate (IPO headwinds, open-weight Chinese competitors offering free access).