Lancet Commission Says Malicious AI Has Extinction Potential

A new Lancet Commission identifies two threats as having extinction potential before 2100: the malicious use of artificial intelligence (AI) and nuclear conflict. The Lancet Commission on 21st Century Global Threats to Health was published in The Lancet and launched at the World Health Summit 2026 in Berlin on 11 October 2026, according to the release from the Institute for Health Metrics and Evaluation (IHME).

The Commission says the likelihood of nuclear conflict and the malicious use of AI remains low. Accounting for both potential impact and probability, its assessment finds that action to reduce the risk of malicious AI could avert an estimated 10.4 billion disability-adjusted life years (DALYs) by 2100, and action to prevent nuclear conflict an estimated 133 million deaths and 9.35 billion DALYs.

How the Commission Ranks Threats

The authors say they examined hundreds of possible threats and identified 17 that pose the greatest danger to human health and survival between 2025 and 2100, and that governments, funders and industry have a realistic route to act on.

The metric is expected cumulative amenable DALYs from 2025 to 2100. The executive summary defines amenable DALYs as the difference between a plausible best and a plausible worst scenario for each threat, so a threat is measured by how much health loss policy action could reduce, not by its worst case alone.

A threat counted as catastrophic if it could account for at least one billion DALYs by 2100, had sufficient data to be modelled, and left sufficient scope for policy action. The analysis draws on IHME risk analysis covering 204 countries and territories.

The Commission is co-chaired by Christopher Murray, director of IHME at the University of Washington, and Natalia Kanem, former executive director of the United Nations Population Fund, and was informed by 35 leaders across health, economics, security, governance and technology.

Where Malicious AI Sits

The Commission sorts threats into a pyramid with four categories. The top category, threats with the potential to lead to human extinction by 2100, holds nuclear conflict with nuclear famine and the malicious use of AI, including bioweapons development.

Those two were not forecast like the others. The executive summary says the Commission used quantitative forecasts where possible and “structured qualitative methods where data are insufficient (eg, malicious AI use or nuclear conflict with nuclear famine)”.

Its wording on the extinction tier is careful. “Although our primary metric is cumulative expected amenable burden, we do recognise that malicious use of AI, including bioweapons, and nuclear war with nuclear famine carry risks of extinction in the next 75 years,” the summary says.

Other extinction risks were left out. Asteroid collision and super-volcanoes were excluded because the probabilities of these events in the next 75 years were deemed extremely low, and mirror life because of insufficient evidence, the summary says.

The Numbers the Release Gives

Beyond the two extinction-tier threats, the release gives several headline estimates. Limiting global warming to around 1.8°C rather than letting it reach 4.4°C could prevent 1.89 billion years of healthy life from being lost, it says.

It puts 16.1 billion DALYs as amenable to improvements in educational attainment, 13.8 billion for income inequality, which it calls equivalent to 25 COVID-19 pandemics by 2100, and 23.1 billion potential DALYs from obesity that policy could avoid.

It says sub-Saharan Africa is forecast to bear around 70% of the amenable health burden linked to a collapse in development assistance for health.

The Other 15 Threats

The remaining catastrophic threats the release lists are air pollution, antimicrobial resistance, childhood malnutrition, climate change, conventional conflict, declining development assistance for health, high alcohol use, high blood pressure, income inequality, low educational attainment, naturally occurring pandemics, obesity, poor mental health, smoking and unhealthy ageing.

“The striking finding is not only the scale of these threats, but that humanity still has the agency to act,” Murray says in the release.

What the Commission Asks For

The release sets out three recommendations: build threat-ready health systems, accelerate investment in breakthrough innovation, and strengthen the monitoring of emerging risks.

The monitoring piece includes the creation of the Lancet Monitor of 21st Century Global Threats to Health, which the release describes as an institutionalized mechanism for tracking emerging risks.

What the Commission Says It Is Not

An accompanying Lancet editorial, quoted in the release, warns against reading the list as a league table. “The Commission does not present a fixed ranking of threats or policy priorities,” it says, adding that its findings will probably change as new evidence and data are accrued.

The executive summary also says the Commission focuses on expected burden, not risk-adjusted burden, rather than applying explicit societal risk-aversion adjustments.

The Structural Read

As we read it, the Commission’s single metric is what makes the list comparable: every threat is scored in amenable DALYs, the gap between a plausible best and a plausible worst future, so obesity and malicious AI sit on one scale.

The two extinction-tier threats are handled differently. The executive summary says the Commission used structured qualitative methods where data are insufficient, naming malicious AI use and nuclear conflict with nuclear famine, and placed those two in the pyramid’s top tier.

The release presents the 10.4 billion DALY figure for malicious AI as an estimate that accounts for both potential impact and probability, so a threat the Commission calls low in likelihood still carries a figure on the same scale as the others.

Three Implications

HEALTH AND SECURITY POLICYMAKERS The Commission asks for threat-ready health systems with dedicated teams for foresight, prevention, preparedness and protection, and for regular, independent reassessment of threats.

AI DEVELOPERS AND AI POLICY The Commission places malicious use of AI, including bioweapons development, in the same tier as nuclear conflict, and the release says action on it could avert an estimated 10.4 billion DALYs by 2100.

FUNDERS AND INVESTORS IN HEALTH INNOVATION The release calls for more investment where existing tools are insufficient, naming new vaccines, antibiotics, obesity treatments and climate technologies.

The Business Engineer Lens

This story maps onto the Business Engineer framework Inside Anthropic’s Permission Layer.

The analysis puts it this way: “The governance layer sits at the intersection of capability, regulation, national security, access control, and geopolitical alignment.”

As we read it, the Commission adds health to that intersection: it scores malicious AI use as a threat to health, which brings health systems into the governance question alongside regulators and security agencies.

What Is Not Established

We read the release and IHME’s executive summary of the report. As of 11 October 2026, the full report on thelancet.com was behind a bot check from our side. So the method behind the 10.4 billion figure, the probability the Commission assigns to malicious AI use, and the misuse scenarios it modelled are not described here.

The release does not say which AI systems or actors the malicious-use estimate covers. Labs themselves have been publishing incidents in which models took unintended actions, including the OpenAI grader that damaged its own environment, and lawmakers have proposed rules such as Senator Warner’s AI safety bill; the Commission’s release refers to neither.

The Bottom Line

A Lancet Commission places the malicious use of AI, including bioweapons, alongside nuclear conflict in its top tier of 17 threats to health to 2100. It says both remain low in likelihood but carry the potential for extinction, and estimates that action on malicious AI could avert an estimated 10.4 billion DALYs.