Trump calls AI risk a hoax, but common sense is no spec
By Nikhil Raghavan · Reporting from San Francisco ·
On Monday, President Donald Trump decided to conduct a policy briefing via speakerphone.
On Monday, President Donald Trump decided to conduct a policy briefing via speakerphone. While Nvidia CEO Jensen Huang was speaking at a summit in Los Angeles, the President called him live on stage to announce that the existential risks associated with artificial intelligence are a “hoax.” According to reporting from CNN and The Washington Post, the President didn’t just dismiss the risks; he linked them to climate change, claiming that the people warning about AI are the same ones who said the world would be extinguished by environmental collapse. He asserted that he has “common sense about AI,” citing his uncle’s tenure at the Massachusetts Institute of Technology (MIT) as a credential, and declared that the only “guardrails” needed are a “STRONG AND SMART (High IQ!) PRESIDENT.”
As someone who spent years on platform-integrity teams and in the weeds of infrastructure, I can tell you that “common sense” is not a technical specification. You cannot "common sense" your way through a reward-hacking failure or a systemic alignment collapse. When the President mocks Dario Amodei, the CEO of Anthropic, for “pretending to be a ‘perfect little angel’” after Amodei called for government oversight, he isn't fighting a political narrative. He is dismissing the very engineers who understand the spec. This is the same rhetorical playbook we saw during the COVID-19 pandemic, where expert warnings about a global threat were brushed aside as political theater until the reality became impossible to ignore.
"Common Sense" is Not a Technical Specification
The administration’s current posture is a masterclass in the "sellable but not shippable" school of governance. Vice President JD Vance and House Speaker Mike Johnson are pushing a “light touch” approach, arguing that “hyper-regulation” would cause the U.S. to lose its competitive edge. This is the strongest version of their argument: that in a geopolitical arms race, the side that iterates fastest wins. They view AI as the “golden goose,” and any friction—whether it's a safety audit or a zoning law for data centers—is a drag on velocity.
But velocity without steering is just a crash in slow motion. The belief that we can simply "out-build" the risk is a fantasy. In my time at AWS, I learned that the things that break at scale don't do so because of a lack of "High IQ" leadership; they break because of edge cases that no one bothered to document. By framing safety as a "hoax," the President is removing the incentive for companies to build robust failure modes. He is essentially saying that the U.S. government will not be the one paged at three in the morning when a frontier model triggers a systemic financial collapse or a critical infrastructure failure.
The China Race is a Poor Excuse for Blindness
The administration is leaning heavily on the China angle, with the President claiming that Xi Jinping has announced China will do “absolutely nothing to stand in the way of AI.” With Xi scheduled for a state visit to Washington next week, the White House is framing AI as a trade chip, mirroring the logic of the Trump-China trade war where tariffs were used as blunt instruments to force concessions.
The danger here is the assumption that China’s lack of guardrails is a competitive advantage. In reality, a catastrophic AI failure in either superpower is a global systemic risk. If the U.S. abandons safety to "win" a race toward a cliff, the prize is irrelevant. We saw this same dismissal of systemic risk during the 2016 United States presidential election, where the vulnerabilities of our digital information ecosystem were treated as fringe concerns until they were exploited in real-time. To suggest that the U.S. can maintain "tremendous CRIMINAL and REGULATORY power" without actual technical guardrails is to mistake a subpoena for an understanding of the code.
Centralizing the Golden Goose
What we are seeing is a modern echo of the First Atomic Energy Act of 1946. After World War II, the U.S. shifted control of atomic energy from military to civilian hands via the Atomic Energy Commission to foster development while maintaining a centralized grip on the technology. The shared mechanism here is the attempt by a political leader to centralize control over a disruptive, world-altering technology while simultaneously dismissing the catastrophic risks highlighted by the scientists who actually built it.
The President isn't proposing a commission; he is proposing himself as the commission. By dismissing the "hoax" of AI risk, he clears the path for rapid hardware expansion—explicitly defending data centers against neighborhood complaints—while ensuring that the "guardrails" are whatever he decides they are on a given Monday. Even Senator John Kennedy’s proposal for a “kill switch” to prevent a “Terminator” scenario, as reported by geekspin, is a superficial solution. A kill switch is a hardware fix for a software alignment problem; it's like trying to stop a flood by closing the front door after the house is already underwater.
The historical pattern is clear: when the state treats technical risk as a political narrative, it loses the ability to regulate the outcome. By replacing technical specs with presidential instinct, the administration is not ensuring American dominance; it is ensuring that when the system fails, we will be the last ones to realize why.