Nuclear war and the malicious use of artificial intelligence rank among the threats to global health: ‘It would inflict an apocalypse’

If the Covid-19 pandemic and the war in Ukraine have demonstrated anything, it is that unexpected, entirely unforeseen and unpredictable events can drastically affect public health and even put humanity at risk. Danger can emerge at any time and from anywhere, but the scientific community has identified a handful of threats with particularly severe potential consequences for human survival.

In an article published on Sunday, a commission of 37 experts convened by The Lancet highlights 17 major risks to global health through 2100: alongside the usual suspects, such as antimicrobial resistance, climate change and malnutrition, new challenges have emerged, including nuclear conflict and the malicious use of artificial intelligence. “Although the likelihood of both dangers remains low, the scale of harm, should they be realized, would inflict an apocalypse on humanity,” the journal warns in an accompanying editorial.

The commission reviewed hundreds of threats facing humanity, but identified 17 that pose the greatest risks to global health and on which governments, the private sector and society can take action to mitigate them. This list of potentially catastrophic challenges includes air pollution, antimicrobial resistance, a naturally occurring pandemic, climate change, child malnutrition, conventional warfare, economic inequality, low levels of education and recent crises such as the rollback of development aid. The list also contains well-known health problems, such as alcohol and tobacco use, poor mental health, obesity, unhealthy ageing and hypertension.

The authors stress that this classification is not fixed. It may change over time, either because of shifting circumstances or new evidence and data. In fact, they note that even today the degree of uncertainty varies substantially across threats. While smoking and malnutrition, for example, have more clearly established trajectories, the malicious use of AI and nuclear war, the only two threats the authors consider capable of causing human extinction, are supported by less established evidence and their long-term course remains unknown.

Uncertainty is high, but the risk is not merely hypothetical. Just a couple of weeks ago, a former employee of AI company Anthropic warned that AI could wipe out humanity “before 2030.” Other researchers in the field subsequently echoed those concerns, pointing to security flaws in AI development, and even the company itself acknowledged that its models carry “existential risks for humanity.” The threat is taken seriously enough that leading technology executives recently warned at the United Nations that AI is “the most important global security issue.”

In its report, The Lancet commission notes that the 17 identified threats extend beyond the realm of health, also jeopardizing global progress, security and stability. The researchers modeled a range of plausible scenarios and, for each one, estimated the years of life lost to poor health that could be averted if action were taken. “Some of the 17 threats are already generating a considerable burden of disease and other health problems, while others — among them, the malicious use of artificial intelligence and nuclear war — could have catastrophic consequences if they materialize,” Christopher Murray, co-chair of the commission, warns in an email response.

The scientist adds, however, that the report’s predictions are not immutable: “What is striking is not only the magnitude of these threats but also that humanity still has the capacity to act against them. The decisions leaders make today will determine the health and well-being of future generations.”

Malicious use of AI

The risk of a catastrophic misuse of AI is “profound,” the authors argue. The report acknowledges a high degree of uncertainty in quantifying the actual danger, but identifies three main pathways through which AI could pose a threat: the development of new pathogens using AI; the integration of AI into critical infrastructure in ways that could trigger cascading failures (for example, hallucinations by AI systems embedded in clinical practice, large-scale political deepfakes or autonomous nuclear weapons); and, finally, the possibility that an autonomous AI system misaligned with the objectives for which it was created could lead to a loss of human control.

According to the study, “this loss of control includes AI systems creating subgoals that lead to the seeking of human disempowerment or elimination, particularly if such agents autonomously gain access to financial networks, key infrastructure, or physical tools in the real world.”

When it comes to quantifying the risk of an AI-related catastrophe, defined as one resulting in the deaths of at least 10% of the world’s population, and the risk of extinction by 2100, the experts consulted for the report are divided. Superforecasters, individuals whose predictions have been shown statistically to be significantly more accurate than those of experts or the general public, estimate the probability of a catastrophe at 2.1% and the risk of extinction at 0.4%. The AI experts consulted, by contrast, put those figures at 12% and 3%, respectively.

Nuclear war

“The devastation that could result from nuclear war cannot be overstated,” the authors warn. The consequences go beyond the immediate death toll and the long-term health effects suffered by survivors. A nuclear detonation would also disrupt supply chains and food production, trigger shortages, undermine government stability and increase the risk of further conflict. In addition, a nuclear war could inject enough particles into the atmosphere to alter global temperatures and reduce agricultural output, a phenomenon known as nuclear famine.

The authors note that between 1948 and 2010 there were an estimated 22 nuclear weapons accidents and 21 incidents in which the world came close to catastrophe. They warn that nuclear crises can arise from escalating international tensions, the accidental detonation of nuclear weapons, terrorism or, for example, the deployment of nuclear weapons through AI systems. Of these scenarios, while accidental launches and terrorist actions remain “serious concerns,” the greatest risk stems from “international crises,” they acknowledge.

This concern is far from trivial given the current upheaval in nuclear geopolitics. For one thing, the U.S. administration has sent mixed signals: six months ago it launched a war against Iran, ostensibly to curb the country’s uranium-enrichment programme and prevent it from developing a nuclear weapon; at the same time, it authorized a nuclear agreement with Saudi Arabia, Tehran’s longstanding regional rival, allowing Riyadh to enrich uranium. Russia, meanwhile, has also revised its nuclear doctrine, expanding the circumstances under which nuclear weapons may be used.

The authors consider the changes in Russia’s nuclear policy “particularly concerning” and call for stronger nuclear non-proliferation treaties and renewed efforts to reduce nuclear arsenals.

Naturally occurring pandemics

The Covid-19 health crisis demonstrated both humanity’s vulnerability and how poorly prepared the world was to respond to a pandemic. But its scale should not have come as a surprise, the experts argue. In fact, it could happen again: while the rate of moderate pandemics has declined over time thanks to improvements in public health, such as epidemiological surveillance and vaccination, the rate of catastrophic pandemics has not followed the same downward trend, in part because of increasing interaction between humans and wild animals.

“Whether an emerged pathogen becomes catastrophic, however, also depends on how widely it spreads,” the commission explains. But several underlying conditions are making that risk more acute. Urbanization, for example, has increased human pressure on animal ecosystems, facilitating cross-species transmission, while also concentrating populations in cities, which fosters more intense transmission of infectious diseases. The hantavirus outbreak that alarmed the world last spring illustrates how easily a zoonotic disease, one that jumps from animals to humans, can spread and trigger a public health crisis.

According to the commission, the threat is real, but its ultimate impact remains uncertain. Among other factors, it will depend on the scientific and technological capacity to develop an effective treatment or vaccine against an unknown pathogen.

Climate change

The commission expanded the number of pathways through which climate change could affect health. These include the direct effects of rising temperatures; the spread of infectious diseases; the direct and indirect consequences of flooding, such as diarrheal diseases; exposure to wildfires and air pollution; changes in food availability; and declines in agricultural productivity, among others. The commission argues that climate change acts as a “threat multiplier,” exacerbating existing socioeconomic challenges.

That said, the commission assumes that the impacts of climate change will not be evenly distributed over the course of the century. Rather, they are expected to intensify as the century progresses and, it adds, “could escalate into an extinction threat in subsequent centuries if emissions are not curtailed.”

Collapse of development aid

Several governments have cut their development aid budgets. The United States froze and dismantled entire programs and, according to an OECD report published this year, funding fell by 23.1%, to $174.3 billion, levels not seen since 2015.

The commission notes that, although the consequences are less severe than those associated with some of the other threats it identifies, the effects of the slowdown in development assistance are concentrated in the world’s most disadvantaged regions.

Unhealthy aging

Life expectancy continues to rise. In 2023, it stood at nearly 74 years and is projected to exceed 78 by 2050. But living longer does not necessarily mean aging well. In fact, the commission notes that the gap between life expectancy and healthy life expectancy has been steadily widening: between 1990 and 2010, each additional year of life expectancy was associated with only one to 10 additional months of healthy life.

The authors warn that this gap is expected to continue growing, with increasing numbers of people spending the final years of their lives burdened by disease and disability.

Quique Bassat, an epidemiologist and director general of ISGlobal who was not involved in the report, welcomes its approach. “It’s an interesting exercise. They didn’t look only at human health but also at the survival of the species,” he says.

According to Bassat, this hierarchy of risks goes beyond the list of potential global health threats regularly published by the World Health Organization (WHO). “The WHO’s lists tend to be very pragmatic and abstract concepts. This is another conceptual framework, but it needs periodic review,” he concludes.

Bassat also highlights the uncertainty inherent in looking 75 years into the future. “You can carry out this analysis in 2026, but perhaps in 2030 a new unknown pathogen emerges that wipes out half of humanity and changes everything,” he reflects. That is why, he stresses, one of the report’s key messages is the need to monitor these threats continuously in order to refine forecasts and reduce uncertainty.

Murray likewise warns that these threats are interconnected and argues that they must be addressed “together, not in isolation.” Beyond monitoring risk, the commission calls for the development of health systems capable of responding to these threats and for accelerated investment in transformative innovations.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition