Lancet Names Malicious AI and Nuclear War as Extinction Threats
A Lancet Commission of 35 global experts warns that malicious AI use could wipe out humanity by 2100. The problem is us, not the machines.
Research · Source: CNN
What happened
A new report in The Lancet just dropped a massive warning. A commission of 35 global experts modeled the greatest threats to human survival by the year 2100. They looked at data from 204 countries and territories. The researchers identified 17 catastrophic dangers spanning disease, climate, and technology. Only two pose a direct risk of human extinction. Those two are nuclear conflict and the malicious use of artificial intelligence.
The researchers measured risk in disability-adjusted life years, or Dalys. A threat is catastrophic if it costs at least one billion lost years of healthy life. They estimate malicious AI could wipe out 10.4 billion healthy years worldwide by 2100. This puts the theoretical damage of AI on the exact same scale as obesity or naturally occurring pandemics. The major difference is that a global AI catastrophe has not happened yet, making the numbers theoretical.
The report emphasizes that the danger comes from how humans use the technology. The threat is malicious use, not a rogue machine. Even AI companies are sounding the alarm. According to the report, a recent initial public offering filing for Anthropic explicitly stated their models could pose a catastrophic or existential risk to humanity. Industry workers have warned AI could wipe us out within a decade.
Key facts
- 17 — Catastrophic threats identified by the commission
- 10.4 billion — Years of healthy life potentially lost to malicious AI by 2100
- 35 — Global experts on the Lancet Commission
- 1 billion — Lost healthy years required to be classified as a catastrophic threat
Why it matters
This shifts the conversation from science fiction to global security. The focus is no longer on rogue machines waking up and deciding to kill us. The medical and scientific community now officially views AI as a tool that could cause extinction-level harm if it falls into the wrong hands. This puts AI governance on the same urgent level as nuclear non-proliferation. The commission co-chairs made it clear that delaying action will only increase future costs.
Expect massive regulatory pressure. When 35 global leaders in health, economics, and security tell the world that your product is a nuclear-level threat, governments listen. The commission is already calling for targeted, coordinated action from world leaders. Builders will face stricter safety mandates, tighter access controls, and intense scrutiny over open-source models. Trust is highlighted as a stabilizing force, meaning public perception will directly impact how hard regulators crack down.
For builders
Expect extreme regulatory friction
Governments will use reports like this to justify heavy-handed regulation. If you build foundational models or deployment infrastructure, expect to pay heavily for compliance and security audits. Open-source builders lose the most as restrictions tighten.
Security becomes a core product feature
The threat is malicious use. Startups that build robust guardrails, threat detection, and misuse prevention tools will see massive demand. Enterprise buyers will pay a premium to ensure they do not accidentally enable bad actors.
Foresight dictates future funding
The commission urges embedding foresight into decision-making. Investors will demand to see your risk mitigation strategies before writing checks. Founders who ignore safety will lose access to top-tier capital.
My take
The real danger is not the machine. It is the human holding the keyboard. When we build powerful tools, we have to assume the worst people on earth will try to use them. If we do not lock down the malicious use vectors now, regulators will step in and shut down the builders.
Original reporting: CNN. This is my rewrite and opinion.