Lancet Commission ranks malicious AI beside nuclear war as a threat that could end humanity

Every threat on the list had to have potential solutions ^[10]. That rule is why malicious AI appears on a health agenda at all, and it puts AI-enabled bioweapons in the same category of policy work as smoking and high blood pressure ^[9][10]. Ibrahim Abubakar, a commission member and professor of infectious disease epidemiology at UCL, said the commission was "starting from the premise that if you are a policymaker today, whether at the global multilateral level or at the national level, these are the potential threats that you could face, and this is the magnitude by which, if you act, you can actually make a difference" ^[11]. The commission measures that magnitude in disability-adjusted life years, or Dalys: healthy years lost to sickness, disability or early death on current trajectories ^[8].

The two extinction-level entries differ from the other 15 ^[17] in the size of the possible loss. "Malicious use of AI, including bioweapons, and nuclear conflict with nuclear famine carry risks of extinction in the next 75 years," the report said ^[2]. An accompanying Lancet editorial was blunter: "The scale of harm, should they be realised, would inflict an apocalypse on humanity." ^[4]

For two of the other threats, the commission put a number on acting. Holding global heating to about 1.8C instead of letting it reach 4.4C could prevent 1.89bn lost years of healthy life ^[12]. A collapse or steep cut in international health aid could cause up to 1.5bn Dalys before 2100, nearly 70% of them in sub-Saharan Africa ^[5]. Roughly 1bn of those lost healthy years would fall in that one region ^[16]. The commission modelled best and worst cases for every threat ^[12]. The Guardian's account of the report gives neither the AI figure nor the realistic route to action the commission says exists for it ^[6].

Christopher Murray, the commission's co-chair and director of the Institute for Health Metrics and Evaluation at the University of Washington, said the process began three years ago with three threats in view, "conflict, climate, and pandemics", and found many more ^[14]. His conclusion was that "governments really need to dedicate some money and staff to the task of anticipation, prevention and preparation for this more diverse set of threats" ^[14].

We think the commission has done one concrete thing for AI misuse this year. It has put the risk in front of governments, next to nuclear war, as a problem with a solution ^[6][11]. The report's horizon runs to 2100 ^[1]. What it prescribes is early action: investment in technological innovation and stronger health systems before risks become crises ^[15]. "Decisions made by leaders today will shape the health and wellbeing of the next generation," Murray said ^[7].