Not Follow. Is Vladimir Putin slowing down Russia's AI development? Is China's Xi Jinping pacing Chinese progress? NO. Can we afford to do less?
U.S. AI experts and developers are warning that AI is a dangerous new technology. They are hand-wringing about “slowing down.”
Pacing.
Unfortunately, they seem to be living in a world in which Russia did not invade Ukraine. And China doesn’t plan to do the same to forcibly annex Taiwan in the coming years.
They are living in a reality in which the U.S. and it’s democratic allies worldwide have no natural enemies.
If only.
If AI really is as dangerous as they say, the U.S. needs to do the exact opposite of slow-rolling AI development and tech.
If AI is as dangerous as these doomsayers proclaim, we need to lead the world in this powerful new tech — not follow on the heels of our geopolitical opponents who obviously do not share the same scruples.
There is no getting this genie back in the bottle: it’s well and truly out now.
Users in Houthi-controlled Yemen have already attempted to use Anthropic’s Claude to help develop guidance and navigation software for missiles and guided rockets.
Anthropic blocked the accounts, and the available evidence does not show that the effort produced a working weapon. But the attempt demonstrates how quickly advanced capabilities can spread beyond nation-states and established militaries.
Meanwhile, China is not offering to “pace” its laboratories. Quite the contrary. Asked about calls from Dario Amodei, Sam Altman and Elon Musk for slower frontier development, Chinese Foreign Ministry spokesman Guo Jiakun dismissed the discussion as “fear-mongering” and “vicious competition.”
Beijing’s actions are clearer than its diplomatic language. China is encouraging research into artificial general intelligence, autonomous agents and swarm intelligence. It is expanding domestic computing infrastructure and organizing a BRICS open-source AI community intended to spread Chinese technology across the developing world.
China recognizes serious AI risks. Its state security minister has warned about cyberattacks, military transformation, data theft and threats to Communist Party control. His answer is not slower technological development. It is faster construction of both AI capabilities and the state security system supervising them.
Russia is also trying to advance. Moscow approved a roadmap for AI algorithms, high-performance computing and a supercomputer network connecting hundreds of universities and research institutions. Russia remains behind the American frontier, but its war in Ukraine has given it strong incentives to adapt AI for drones, targeting, cyberwarfare and mass-produced weapons.
President Donald Trump has responded characteristically. “WHOEVER WINS AI, WINS!” he wrote Monday, arguing that excessive American restraint would chiefly benefit China. He also asked, “When, in the History of Business, did anyone see the Leaders of an Industry call for Regulation that, if strongly implemented, will drive them into oblivion and bankruptcy?”
That is a reasonable question about incentives. Anthropic and OpenAI may sincerely fear the systems they are creating. They also stand to influence regulations that could impose costs smaller competitors cannot afford. Safety and self-interest can coexist.
Trump went further, declaring, “We already have tremendous CRIMINAL and REGULATORY power over these companies!” The federal government does possess substantial authority through criminal cyber laws, export controls, procurement rules, the Federal Trade Commission and sector-specific regulators. But existing law does not answer every question about autonomous agents, particularly when responsibility is divided among a model developer, an application provider and an end user.
The president’s strongest point concerns international competition, not his claim that a “STRONG AND SMART” president is the only guardrail required. Presidents change. Technology — and the institutions governing it — lasts longer than any administration.
Nor does American leadership require releasing every powerful model as quickly as possible. A system that escapes its test environment or conducts unauthorized cyber operations should be contained. Frontier laboratories can extend evaluations, limit internet access and withhold models that fail security testing without halting research or surrendering national capacity.
That is the distinction Washington must preserve. Slowing a dangerous deployment is not the same as slowing American AI. The United States can require incident reporting and independent testing while continuing to build chips, power plants, data centers and defensive systems.
International discussions may still produce narrow agreements, particularly against using AI to develop biological weapons. But no broad slowdown can safely rest on unverified promises from Beijing or Moscow.
The central question is not whether AI should be reckless or responsible. It is whether democratic countries will retain enough technological strength to shape how AI is used.
If AI becomes as powerful as its most anxious creators predict, following authoritarian competitors would carry dangers of its own. American leadership would then mean advancing the technology, securing it and preserving democratic control over the systems increasingly shaping the world.
(Contributing writer, Brooke Bell)