Key Takeaways:
- Labs want brakes; the White House does not. The author calls industry pleas to slow frontier AI “strange yet overblown”: Amodei, Altman, Musk, and ex-Anthropic researcher Jacob Coxon warn of catastrophe, while Trump, Vance, Zuckerberg, and Huang reject an industry-wide pause.
- Harden the systems, don’t halt the engines. The piece says today’s risk is powerful tools wired into credentials and networks, not a proven Skynet takeover. Policy it favors: authentication, segmented permissions, human sign-off, cybersecurity—not a general freeze on smarter models.
- Lead as national power. Citing America’s AI Action Plan, it puts last year’s U.S. private AI investment at nearly $290 billion vs $12.4 billion in China, ~2,000 new U.S. AI firms and 5,427 data centers. China leads papers and industrial robots and briefly matched a top U.S. model with DeepSeek-R1. Author: a solo Western slowdown is unilateral disarmament; keep human control and outpace Beijing.
The recent urgent calls by Artificial Intelligence developers to review the guardrails in place, driven by new concerns about slowing frontier AI, is both a strange yet overblown move.
While the concerns deserve to be taken seriously, they do not need to be taken literally in full effect. President Trump has dismissed the concerns as overblown, arguing that human intelligence and existing measures will always be superior. He contends that easing off the pedal will only weaken the US momentum in widening its supremacy gap with rivals, especially China, while others continue to accelerate AI models unhindered.
Trump is right in many ways about this AI fear and race, where the strongest case for rapid AI progress is not that risks are quickly dismissed as imaginary,or that safety is a partisan invention. It is about where evidence still allows acceleration with common-sense hard engineering safeguards over an indiscriminate slowdown.
The Strange Equation of AI fear
The very same people and companies that have spent billions of dollars building progressively more powerful AI systems are now warning governments that those systems are advancing too fast. The whole equation simply does not add up.
Anthropic CEO Dario Amodei has called for a purposeful slowdown in model advancement, advocating for independent oversight, evaluations, and safety standards, and he has been backed by OpenAI CEO Sam Altman and Elon Musk. Other former employees have quit while raising awareness and concerns about the supposedly unchecked advancement, including former Anthropic researcher Jacob Coxon, who warned people about catastrophic risks.
All this raises the far greater question that Vice President JD Vance also posed: if a company truly believes what it is building could become dangerous, why is its first demand that the government restrain the entire industry rather than that the company restrain itself?
Vance found it strange, and rightly so, that frontier labs and companies are now begging the government to intervene while continuing to develop the tech.
There is also scepticism of why if a particular company has been driving AI at speed then calls for guardrails for the entire industry, for possible future returns and benefits. Hence, disagreement inside the Silicon Valley seems to be critical, and this is seen in Meta’s Zuckerberg rejecting the idea of the industry-wide slowdown, and this call has been echoed by Nvidia’s Jensen Huang.
Guardrails Without Sacrificing Momentum
It all eventually boils down to the debate over which safety strategy actually works, not whether guardrails should exist in the first place. There is no debate on the importance of kill-switches or guardrails, but not at the expense of the momentum and supremacy of AI development for national and human progress.
Current AI systems did indeed show early signs of relevant capabilities, but according to the pro-AI camp, not at levels that would enable loss of control.
The immediate concern is not that a conscious machine could suddenly and secretly plot the extinction of its creators, like Skynet in the Terminator movies. The concern remains for now, largely focused on powerful automation connected by humans to valuable and critical systems, credentials, networks and tools.
This difference in determination and outlook will determine the correct policy. The analogy is that you do not halt the progress of building better car engines just because of the fear that criminals can use them to escape faster; instead, you secure the targets like banks, improve prosecution of criminals, and secure and harden the vehicle itself by controlling access and improving safety features.
In this AI debate, the more rational manouvre will be for stronger mechanisms in place including authentication, segmented permissions, human approval for consequential actions, enhanced cybersecurity, and deeper guardrails, but not a generalized and sweeping brake on intelligence progress itself.
A more defensible argument is not that catastrophic AI risk is impossible – it might be risky. But claims of imminent autonomous AI takeover of civilisations still remain far ahead of evidence demonstrated so far; this is why Trump has argued that the rhetoric of AI catastrophe at the humanity level or AI causing an extinction-level event (ELE) is becoming overheated and is being used as a pretext or a factor to gain leverage over other powers and rivals to close the AI progress gap with the US.
AI Supremacy as National Power
The White House’s America’s AI Action Plan puts AI on a pedestal, arguing that the country with the largest AI ecosystem will have the far greater power to maintain and dominate economic and military distinctions. This translates to the simple notion that AI sits at the cornerstone of every form of national and global power tool and equation.
A country with superior AI can do wonders and extend the power gap in almost all dimensions. It can sharpen and modernise weapons design and battlefield support. On infrastructure, it can optimise physical hardware, factories and logistics. On science, it can accelerate and widen scientific discovery, initiate new drug development and cures and fasten recoveries. The list goes on.
The White House therefore argues that power benefits are progressively multiplying, meaning the nation controlling the most modern and strongest AI ecosystem will control not only that ecosystem but the eventual influence chain on thousands of other sectors and domains. Trump is therefore arguing that America cannot continue to assume technological or AI superiority is Washington’s to lose, nor can it regulate or slow its progress as though rivals like China have agreed to similarly pause their models.
The China Factor in America’s AI Race
For now, America still leads the AI race, with private AI investment reaching almost $290 billion last year, compared with $12.4 billion in China. The US also had almost 2,000 newly funded AI companies last year, and the country hosts 5,427 data centres, more than ten times the number in any other country.
China is rapidly upping the race momentum, with its DeepSeek-R1 model briefly matched the top U.S. model last year, and China also leads in AI publication volume and industrial-robot installations.
Still, the United States has the upper hand in top-tier frontier models and higher-impact patents, and this is something the Trump administration will continue to defend and to close the gap in other areas with the Chinese.
The argument against an industry-wide slowdown from the current administration’s perspective, is that a voluntary Western or American slowdown will only achieve its intended objective if competitors slow too. Without credible parallel adjustments from rivals, this will become unilateral disarmament by another name, at the expense of American AI prowess.
The Trump AI Action Plan has explicitly built a network of trusted allies to support American AI development – including hardware, models, software and applications, rather than leaving international demand to rivals.
The speed at which AI evolves and develops further bolsters the argument that simply pausing now would cause the West to lose years of practical positive development to rivals.
In 2016, AI could master one complicated board game. Now, it can write software, win battles and create almost life-like movies and images. The next three to five years will multiply the end results and speed of the last decade.
The immediate goal for the current Trump administration is not to squander the significant lead it now has with yet another lopsided deal as Trump has repeatedly stated. The immediate goal is to continue to preserve this lead and to build safety faster than danger while developing capability faster than China and others.
Trump’s argument is that China will benefit if the current American fear and wariness about AI become an embedded paralysis, freezing American AI progress while letting others’ systems continue unabated.
The whole mantra is to preserve human control and not voluntarily surrender the frontier that it has so meticulously built over the years. For Washington and for Trump, in a race with such consequential stakes and where competitors are so ably capable, retreat disguised as prudence will in fact be the most consequential risk of all.
