Lancet Commission Identifies Malicious AI and Nuclear War as Potential Existential Threats

A new report from The Lancet Commission on 21st Century Global Threats to Health has identified the malicious use of artificial intelligence and nuclear conflict as two threats with the potential to cause human extinction. Published on October 10, 2026, the assessment examines risks that could affect human health, economic stability and global security over the remainder of the century. The commission brought together international experts who evaluated hundreds of possible threats before identifying 17 that require urgent attention from governments, researchers and policymakers. Although the likelihood of an extinction-level event remains uncertain and is considered low, the report warns that the consequences could be so severe that preventive action is essential. The findings were presented in connection with the World Health Summit 2026 in Berlin.

Artificial intelligence is highlighted because of the possibility that increasingly capable technologies could be deliberately used to cause widespread harm. The report raises concerns about malicious actors exploiting AI to support the development of biological weapons, facilitate cyberattacks or disrupt critical infrastructure. Advanced AI systems could also increase the speed and scale at which dangerous information is generated and used. However, the commission’s assessment does not mean that AI itself will inevitably cause human extinction. Instead, it identifies malicious use as a potential pathway to catastrophic outcomes and argues that governments and technology companies should strengthen safeguards, monitoring and international cooperation to reduce the risks. The report places AI safety alongside other major global security concerns that require sustained attention.

Nuclear conflict is the second threat identified as having extinction potential. The commission warns that a large-scale nuclear war could cause immediate mass casualties while producing longer-term consequences through environmental damage, disruptions to agriculture and widespread food shortages. A nuclear winter, in which smoke and soot reduce sunlight reaching the Earth’s surface, could severely affect food production across multiple regions. The report estimates that action to prevent nuclear conflict could avert approximately 133 million deaths and 9.35 billion disability-adjusted life years by 2100. These figures are model-based estimates of potential health losses avoided, rather than predictions that a nuclear war will occur. They nevertheless illustrate the enormous human consequences associated with a major nuclear confrontation.

Beyond AI and nuclear weapons, the commission identified 15 additional threats that could undermine human health and survival. These include climate change, air pollution, antimicrobial resistance, naturally occurring pandemics, childhood malnutrition, high blood pressure, obesity, smoking, excessive alcohol consumption and poor mental health. Other concerns include income inequality, low educational attainment, conventional armed conflict, declining development assistance for health and unhealthy ageing. The report emphasises that these threats are interconnected rather than isolated problems. For example, climate-related disasters can damage health systems, deepen poverty and increase food insecurity, while conflict can interrupt vaccination programmes and access to essential medicines. Addressing these risks therefore requires governments to coordinate health, environmental, economic and security policies rather than treating each threat separately.

To reduce these dangers, the commission recommends building health systems that can respond to major emergencies, accelerating investment in scientific and technological innovation, and improving the monitoring of emerging threats. Governments are encouraged to strengthen international cooperation, prepare for new infectious diseases and develop safeguards against the misuse of powerful technologies. AI developers also face pressure to improve security testing and prevent their systems from being exploited for harmful purposes. The report estimates that reducing risks associated with malicious AI could avert approximately 10.4 billion disability-adjusted life years by 2100, although this is a modelled estimate rather than a guaranteed outcome. Ultimately, the commission argues that the future is not predetermined: coordinated decisions made by governments, businesses and researchers could significantly reduce the likelihood and impact of catastrophic.