The Lancet Put Malicious AI Next to Nuclear War. Your Diet Is Still the Bigger Threat.

A major Lancet report has put “malicious use of artificial intelligence” on the same short list as nuclear war: the only two threats this century that the authors say carry real potential to end humanity. That is a striking line to see in one of the world’s oldest medical journals, and it has produced a wave of headlines suggesting we have until 2100 to live.

The actual report is less cinematic and more useful. As UNILAD Tech reports, the Lancet Commission on 21st Century Global Threats to Health identified 17 catastrophic threats to human health between 2025 and 2100, each big enough to cost more than one billion years of healthy life. UNILAD frames it as “do we only have 74 years left?” The Commission says nothing of the kind. 2100 is simply where the model stops counting, not a predicted end date.

What the report really measures is preventable disease burden. And by that yardstick, the biggest threat on the list is not a rogue AI. It is the stuff on your plate.

What the Commission actually found

The report was published in The Lancet in October 2026 and launched at the World Health Summit in Berlin on October 11. The Commission screened hundreds of candidate threats (coverage cites figures of 300 or more), drawing on risk modeling for 204 countries and territories from the Institute for Health Metrics and Evaluation (IHME) in Seattle.

To make the final list, a threat had to clear three bars: it could plausibly cause at least one billion DALYs by 2100, there had to be enough data to model it, and there had to be meaningful room for policy to change the outcome. A DALY is one year of healthy life lost, whether to early death or to living with illness or disability.

The headline figures use a narrower measure the authors call “amenable” DALYs. For each threat, the team modeled a plausible best-case and a plausible worst-case future. The gap between the two is the amenable burden: the health that policy, money and innovation could still save. That makes the numbers a map of where effort pays off, not a forecast of how many people will die.

The Commission also stresses what the list is not. It is not a ranking of policy priorities, and some huge killers, including ischemic heart disease and several cancers, were left off because they already get heavy attention and funding. Their absence, the authors say, should not be read as a sign they matter less.

A note on the chart: the Commission’s per-threat numbers are in the paper, but only some have been widely reported. We plotted the eight figures we could confirm across the Commission’s release and press coverage, to scale. The climate number is especially conservative. It covers only direct, measurable harms from heat and air pollution, and the authors say it leaves out displacement, drought and mental health effects.

The AI threat: what the Lancet means by “malicious use”

For a tech audience, the key word is malicious. The Commission is not warning about chatbots giving bad medical advice or an AI spontaneously deciding to wipe us out. Its concern is people using AI to cause mass harm. The pathway most coverage highlights is AI-assisted design of biological weapons, alongside attacks on critical systems such as health care and finance.

According to the IHME release, “while the likelihood of nuclear conflict and the malicious use of AI remain low,” the two “present consequential threats because of their potential for extinction.” A Lancet editorial put it more bluntly, as quoted by Common Dreams: “Although the likelihood of both dangers remains low, the scale of harm, should they be realized, would inflict an apocalypse on humanity.”

The 10.4 billion DALY estimate for AI is the softest number in the report, and the authors know it. There is no historical dataset of AI catastrophes to model from, so for both AI and nuclear conflict the Commission used what it calls structured qualitative methods and expert judgment. Health Policy Watch reports the uncertainty range for the two extinction-tier threats runs from zero, meaning no harm at all, up to 15.6 billion and 24.5 billion DALYs. Coverage of the report also cites forecasts putting the chance of AI-driven human extinction by 2100 somewhere between 0.4% and 3%, depending on whether you ask generalist superforecasters or AI specialists.

There is also a quirk of the metric that cuts the other way. DALYs count healthy years lost by people who exist. An extinction event erases every future generation too, and none of that shows up in the tally. So the AI and nuclear numbers are simultaneously very uncertain and, by design, a floor on what an actual extinction would cost.

If this sounds abstract, the building blocks are already showing up in security research. We have already covered research into self-replicating AI worms aimed at smart home and IoT gear, and India’s CERT-In has issued a warning about rogue AI agents across multiple providers. None of that is extinction-grade, but it shows why the Commission treats misuse, rather than machine rebellion, as the realistic tech risk.

All 17 threats, in one table

The other 15 threats span chronic disease, social and economic conditions, the environment and conflict. Here is the full list as published, with our short gloss on what drives each one.

The boring threats are the biggest ones

Strip away the extinction headlines and the report’s sharpest finding is how unglamorous the top of the list is. Obesity, poor diet and low physical activity together carry about 23.1 billion amenable DALYs, more than twice the AI estimate. Coverage of the report says it is the leading threat in 154 of the 204 countries studied. Low educational attainment (16.1 billion) and income inequality (13.8 billion) come next, ahead of both doomsday scenarios.

Health Policy Watch reports that smoking, poor mental health, unhealthy ageing and conventional conflict each top 7 billion, while childhood malnutrition sits at the bottom of the range at 1.02 billion. The Commission attributes about 3.44 billion amenable DALYs to antimicrobial resistance.

That ordering is exactly what worries some observers about how the report is being covered. Writing in South Africa’s BusinessDay, political scientist Anthony Butler argues that AI extinction fears risk overshadowing preventable health threats, and that mundane problems amenable to policy could get pushed down the agenda by apocalyptic framing. Others flag that the extinction-tier numbers are, in effect, structured opinion rather than measured event rates, which makes them hard to compare directly with obesity or smoking data that span decades.

What the co-chairs say

Murray, IHME’s director and a professor at the University of Washington, has framed the findings around agency rather than doom. “The striking finding is not only the scale of these threats, but that humanity still has the agency to act,” he said, as quoted by ScienceAlert. He has also argued that delay only shifts and increases the costs into the future, and that the threats are interlinked and need to be tackled together.

Kanem, who previously led the United Nations Population Fund, focused on the cost of waiting. “Recent global events show us just how quickly stability can give way to crisis, and how costly it is to respond only after disaster strikes,” she said in the Commission’s release. She also warned that these threats risk amplifying existing inequalities, with the heaviest burden falling on the most vulnerable communities. Health Policy Watch quotes her describing a “non-zero risk of human extinction during this very century” from the two top-tier threats.

Three things the Commission wants leaders to do

The policy section is short and practical, organized around three verbs.

• Build readiness. Create threat-ready health systems with dedicated foresight, prevention, preparedness and protection teams, backed by durable financing rather than ad hoc crisis spending, and designed to reach lower-resource and marginalized communities.
• Back breakthroughs. Invest in innovation where existing tools fall short, including new vaccines, antibiotics, obesity treatments and climate technologies, with governance and equitable access built in.
• Track threats. Reassess the list regularly and independently. The Commission is setting up a Lancet Monitor of 21st Century Global Threats to Health to follow emerging risks over time.

That last point matters for the AI entry in particular. A threat estimate built on expert judgment in 2026 is a snapshot. Capabilities, safeguards and regulation will move fast, and a standing monitor is the Commission’s way of admitting the number will need revisiting. The same AI tools raising alarms are also producing gains in medicine, from drug discovery to models that estimate cancer survival from a face photo, which is partly why the report targets misuse rather than the technology itself.

The takeaway

The Lancet Commission did not set a doomsday clock. It produced a ledger of preventable harm, and AI earned a line on it because a small chance of catastrophe multiplied by an enormous downside adds up to a big number. For the tech industry, the message is specific: the risk regulators and health experts are worried about is people weaponizing AI, especially in biology. For everyone else, the bigger bill is still coming from diet, schooling and inequality, and those are problems we already know how to work on.

Sources and further reading

• UNILAD Tech: Leading scientists identify 17 catastrophic threats to humanity with 2100 warning
• IHME: New Lancet Commission provides roadmap to protect economies, security and global stability
• EurekAlert: The Lancet press release on the Commission
• The Lancet: Commission on 21st century global threats to health
• The Lancet: Rethinking threats to human survival (comment)
• IHME: 21st century threats to health overview
• Health Policy Watch: Is the world ending? A Lancet Commission says maybe
• Euronews: Nuclear war and malicious use of AI could end humanity by 2100, scientists warn
• ScienceAlert: Experts identified the gravest health threats facing humanity
• Common Dreams: Malicious use of AI, nuclear war pose extinction threat
• BusinessDay: Anthony Butler on AI extinction fears and preventable health threats
• Juta MedicalBrief: Lancet Commission warns of 17 dire threats to global health
• Nature World News: Malicious AI and nuclear famine carry low-probability extinction risk

About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.