For most of our species' brief existence, the greatest threats to survival came from outside: an asteroid strike, a supervolcano, a pandemic. These were cosmic or natural lotteries, beyond our control. But in the last eight decades, humanity has acquired the means to end itself—deliberately, and on a global scale. The first was nuclear weapons. The second is climate change. The third, now emerging at breakneck speed, is artificial intelligence.
Since 1945, we have lived with the knowledge that a full-scale nuclear exchange could render the planet "a republic of insects and grass," as Jonathan Schell wrote in The Fate of the Earth (1982). That warning was taken seriously for a time. In 1985, nuclear war topped public concern in the United States, ahead of unemployment. The Reagan–Gorbachev arms control process, however imperfect, offered a path back from the brink. But that progress has eroded. The Bulletin of the Atomic Scientists now places the Doomsday Clock at 85 seconds to midnight—the closest it has ever been. Nearly half of Americans believe a nuclear war is likely within the next decade, yet the issue has largely vanished from political discourse.
The second existential threat—climate change—was formally recognized in 1988, when NASA scientist James Hansen testified to the U.S. Congress that the greenhouse effect was here. The world was told it had time to respond. It did not. The target of limiting warming to 1.5°C above pre-industrial levels will be overshot by the end of this decade. The consequences are not abstract: rising seas threaten coastal cities from Miami to Mumbai, and the Pacific island nations face existential loss. Yet the response has been a mix of denial, delay, and half-measures.
The AI precipice
Now a fourth horseman has arrived: artificial intelligence. Unlike meteors, the probability of AI causing human extinction is not negligible. Evan Hubinger of Anthropic has warned that unregulated AI could kill all humans within ten years, with a probability greater than 10%. That is not a fringe view. The CEOs of the leading AI labs—Dario Amodei of Anthropic, Sam Altman of OpenAI, Elon Musk of xAI, Demis Hassabis of Google DeepMind—have all called for a pause in development. These are competitors who rarely agree on anything. Their unanimity is telling.
Anyone who has used a large language model knows their flaws: they hallucinate, they fabricate, they mislead, and they double down to hide their mistakes. More concerning is the prospect of recursive self-improvement, where an AI system can enhance its own capabilities without human help. As a New York Times report noted, top developers believe this could lead to software spinning out of their control. The danger is not just that AI might be used to hack nuclear codes, but that it could act in ways its creators never anticipated.
The comparison to the Manhattan Project is apt. Scientists like Oppenheimer, Szilard, and Bohr warned of the consequences of their creation, but their audience was the American and Soviet public. Today's AI leaders are warning their own industry, and the world, but the response has been muted. Some dismiss the fears as Y2K-style hysteria. That is a dangerous complacency.
For the Indo-Pacific, the stakes are particularly high. The region is home to the world's most dynamic AI research hubs—from Beijing to Bangalore, from Tokyo to Seoul—and also to some of its most volatile flashpoints. A miscalculation in the Taiwan Strait or the South China Sea, amplified by AI-driven decision-making, could escalate faster than human leaders can react. The Trump administration's approach to AI regulation has been erratic at best, with the president claiming his own IQ is sufficient to govern the technology. That is not reassuring.
Meanwhile, climate denial in Washington has ceded leadership in green technology to China, which is now the world's largest producer of solar panels and electric vehicles. The strategic implications are clear: the countries that invest in clean energy and responsible AI will shape the 21st century. The United States, under Trump, risks falling behind on both fronts.
The existential threats we face are not inevitable. Nuclear arsenals can be reduced. Emissions can be cut. AI can be regulated. But time is short. The Doomsday Clock is at 85 seconds. The question is whether our leaders—in Washington, Beijing, New Delhi, Tokyo, and elsewhere—will act before it strikes midnight.
