On a warm Texas night, after the noise of the Republican National Convention had faded and the motorcade was ready to leave Dallas, President Donald Trump found himself fielding a question that sounded less like politics and more like science fiction. Reporters, clustering on the tarmac, asked whether he was concerned about artificial intelligence eventually wiping out humanity. The president waved off the doomsday scenario with characteristic impatience. That wasn’t the thing keeping him up at night. What worried him, he said, was something far more concrete and, in his mind, urgent: the race against China. If the United States lost the artificial intelligence race, he warned, the country would be placed in a very bad position. He then offered his own assessment of the scoreboard, claiming America was leading China by what he called a pretty good margin and that in this field a year was a lot. It was the sort of confidence Trump often projects on matters of national competition, a blend of optimism, instinct, and a desire to project strength. But the exchange was more than a simple sound bite. It captured a deep divide in how the people closest to AI understand the future. On one side sit many of the researchers who are actually building the technology, some of whom privately and publicly argue that AI might kill us all within a decade. On the other side sits a president who, when confronted with that apocalyptic warning, chose instead to focus on the rivalry with Beijing. He even mentioned that Chinese President Xi Jinping would visit Washington later this year, and said he expected the meeting to be great. But he made no mention of putting AI safety on the agenda. For Trump, the conversation about AI was, above all, a conversation about winning.
To understand why that answer felt so jarring, it helps to understand what has been happening inside the AI research community in the weeks before Trump spoke. A former researcher at Anthropic, Jacob Coxon, had just quit his job over exactly these fears. In a series of posts on X, Coxon did not limit himself to technical critiques. Coxon had worked on pretraining at both OpenAI and Anthropic, meaning he had seen the raw ingredients of these systems up close. He accused the two leading AI companies, OpenAI and Anthropic, of acting irresponsibly by racing straight to self-improving superintelligence and gambling with our lives. It was a blunt, angry warning from someone who had spent time on the inside, working on the very systems that he now feared. Coxon said the people building these systems earnestly believe that AI could kill us all by the end of the decade, and he insisted that this was not a marketing stunt. That last part mattered, because it is easy to dismiss AI fears as science fiction or as a way for tech companies to seem dramatic. Coxon was saying the opposite: that the engineers and researchers closest to the work are not casually speculating. They are genuinely terrified. His message was then amplified by Evan Hubinger, a scientist who leads alignment work at Anthropic. Hubinger said his colleagues really do believe AI could kill all humans and that he personally puts the odds at more than one in ten within the next decade. For anyone not steeped in AI jargon, alignment is the challenge of building intelligent machines that act in accordance with human values rather than finding clever ways to do things we never asked for. It is a hard problem, and honestly acknowledging the danger from inside a company whose entire purpose is advanced AI made the warning harder to ignore. The posts did not read like academic abstractions. They read like people who had seen the engine under the hood and had begun to suspect it could not be safely driven.
Once those warnings entered the public timeline, the political reaction was almost instantaneous, and surprisingly bipartisan. The usual lines between left and right blurred, for a moment, because the future of the species is not a typical partisan issue. Senator Bernie Sanders of Vermont, who has long pushed for stricter AI regulation, went straight to the most radical proposal available. He tweeted that he planned to introduce legislation to ban superintelligence and pause AI development. In his view, the very people building this technology were admitting that it could threaten the future of humanity, and he thought that should be enough to stop the race before it reached the finish line. On the Republican side, Florida Governor Ron DeSantis also weighed in, writing on X that it was alarming to think these technologies could elude human control. He added that technology should enhance the human experience, not supplant it. The phrase was a reminder of a deep, almost philosophical anxiety: not just that AI might be used badly, but that it might become something that no one controls at all, a force capable of replacing human agency rather than serving it. Senator Chris Murphy, a Connecticut Democrat, saw a simpler problem and a simpler solution. He pointed to Coxon’s post and argued that the blind race to build a death machine first could be easily solved through a regulatory system that allowed research to continue, but not in a way that destroys us. That word, easily, was probably too optimistic, but it captured the frustration of many on Capitol Hill who think the answer is obvious: let the science move forward, but build guardrails as you go. The fact that such divergent voices could agree on the basic seriousness of the threat, if not the exact response, showed how much the conversation had shifted.
Perhaps the most revealing reaction came from Senator Ted Cruz of Texas, who appeared on The View to talk about the same cascade of warnings. Cruz acknowledged the concern and said he was working with Senate colleagues on legislation that would address catastrophic risks from AI. That sounded like a responsible, safety-focused position. But then he went on to explain why he was not entirely willing to embrace the idea of slowing down. China, he said, was going full speed ahead, and whatever the United States did, China was not going to stop. In Cruz’s mind, that changed the moral calculation. If AI was going to be dangerous, he said, it was better for America to be the one leading it. His exact words became an instant headline: if there are gonna be killer robots, I’d rather they be American killer robots than Chinese killer robots. The sentence was strange enough to feel like a parody, but he meant it, and it reflected a broader logic inside Washington, a logic that also happened to echo the president’s own tarmac comments. The existential threat was real to Cruz, but the geopolitical threat was more real. He did not want to win an extinction race, but he definitely did not want to lose one. That framing has become one of the central tensions of the entire AI debate. Most experts agree that unfettered competition could be dangerous, and many safety advocates have suggested that the United States and China should coordinate to slow down the pace of development until proper safety guardrails are established. But in a world where the two superpowers already distrust each other, calls for cooperation often die quickly. The killer robot preference was an extreme version of a very common political instinct: it is better to be number one in anything, even if the thing itself is terrifying.
Set against all this, Trump’s dismissal of AI extinction risk was not exactly a fringe position within his party, but it did represent a particular way of thinking about technology and power. The president seemed to believe that the real danger was not what AI might do to all of humanity, but what it might do to America’s standing in the world. In that vision, the human race is not the endangered species. The United States is. And the enemy is not a machine that escapes human control, but another nation that beats us to the finish line. This is not an unreasonable political instinct. For decades, American leadership in technology has been bound up with national security, economic growth, and cultural influence. Being first has been almost a synonym for being best. But AI is different from nuclear weapons, semiconductors, or even the internet, because it raises the possibility that the tool itself becomes the rival. Some researchers and safety advocates have suggested that the only way out of this bind is for the United States and China to cooperate, slowing the race long enough to establish safety standards. That idea, however, runs into the same wall every time: neither side wants to be the one that blinks. This tension is not just an American problem; it is a global one, but the most important decisions are still being made in two capitals that can barely agree on anything else. Trump’s refusal to even mention AI in the context of the upcoming visit from Xi Jinping suggested that, at the highest level of the American government, AI safety was not yet a diplomatic priority. The race was being run because it was being raced, and the notion of pausing to check the brakes seemed almost un-American to the leaders who wanted to win. It is a strange moment in history. The technology that could save us, transform medicine, energy, and every other field, is also the technology that some of its own creators believe could erase us. And the people we elect to keep us safe are arguing about which country will get there first.
In the end, the exchange on that Dallas tarmac was a mirror for the entire AI debate. Trump was asked about the possibility of human extinction, and he answered with a comparison to China. That small movement, from apocalypse to rivalry, is exactly how public conversations about AI often go. The claims from researchers like Coxon and Hubinger are genuinely alarming. They are not talking about job displacement or fake videos. They are talking about a technological threshold beyond which humans might not be able to stop what they have built. If there is even a 10 percent chance of that happening within the next decade, it should be treated as a crisis on a scale that dwarfs nearly every other issue in politics. Instead, the political system is doing what political systems do: turning it into a contest. Sanders wants to pause everything. DeSantis wants to make sure technology never replaces people. Murphy wants a regulation that somehow lets research continue without destroying us. Cruz wants American killer robots, because at least they would be ours. It is tempting to mock the absurdity of an American preference for killer robots, but the deeper issue is that no one has found a way to be both safe and first. And the president, standing in the dark with the noise of the convention behind him, wants to win. None of these answers is truly satisfying, and that is probably because the question itself is too big for a sound bite. The humanization of this story is not found in the policy details, but in the uncomfortable fact that the people building AI are afraid of their own creation, and the people governing the world are still stuck in an older, simpler fight. We want to believe that the adults in the room know what they are doing. Yet one of the adults in the room just quit his job because he was too scared to keep working on it at full speed. Another adult, the president of the United States, looked at the threat of extinction and said the real problem was losing to China. Until those two realities can be brought together, the race will continue, the researchers will keep warning, and the rest of us will be left to wonder whether winning the future means surviving it. That is the question that no one, on any tarmac, has fully answered.
