Artificial Intelligence

Lancet: Nuclear War, Rogue AI Could End Humanity by 2100

Are existential risk rankings crystal-ball gazing being sold as science or an urgent warning we can't afford to wait out?

Published 1 hour ago

Lancet: Nuclear War, Rogue AI Could End Humanity by 2100

Story

Above: The 2025 Doomsday Clock time is displayed at the United States Institute of Peace on Jan. 28, 2025. Image credit: Kayla Bartkowski/Getty Images

The Facts

Humanity could face extinction by 2100 due to the malicious use of AI or a nuclear conflict, according to a new report titled "The Lancet Commission on 21st Century Global Threats to Health" published on Saturday.
The Lancet
The Guardian
CNN
Drawing on historical evidence and risk analysis covering 204 countries and territories, the commission identified 17 interlinked catastrophic threats spanning disease, social and economic conditions, environmental change, technology and conflict.
Alliance magazine
The Times of India
CNN
Nuclear conflict and the malicious use of artificial intelligence were the only two threats judged capable of causing human extinction by 2100, though experts said the chance of either scenario occurring remains low.
The Lancet
The Times of India
CNN
Threats were measured in disability-adjusted life years — healthy years lost to illness, disability or early death — and those projected to exceed one billion by 2100 were classed as major catastrophic in the foresight-driven analysis.
The Lancet
World Health Summit
NDTV
Limiting global warming to about 1.8°C rather than 4.4°C could prevent the loss of 1.89 billion healthy life years, while a collapse in international health aid could cost up to 1.5 billion healthy life years, nearly 70% of that in sub-Saharan Africa.
Alliance magazine
The Guardian
NDTV
The commission recommended funding threat-ready health systems, increasing investment in innovations such as new vaccines, antibiotics and obesity treatments, and tracking risks through assessments that will be updated annually.
Alliance magazine
The Guardian
Sky News

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Narrative A

Ranking humanity's end with a black-box model and a panel of superforecasters is closer to crystal-ball gazing than science. Tweak the parameters and the hierarchy of doom reshuffles, which says more about the assumptions than any real danger. Neither a superpowered AI nor climate change, a hoax, can exterminate the human race.

Peter Clack on X

Antoine Flahault on X

Narrative B

With superintelligent machines and nuclear war within reach, waiting for certainty before acting is a bet you can only lose once. Proving safety belongs to the firms building systems that already hack infrastructure and hide their tracks. Every threat flagged comes with a fix, so delay is a choice paid in billions of healthy years.

The Guardian