AI, China, and the False Choice Between War and Surrender

What Happened

On Saturday, Dario Amodei, CEO of Anthropic, published an essay warning that "a Chinese lead in AI would pose grave danger for the United States and the world." He called for the U.S. to maintain chip export restrictions to China and urged a global slowdown in AI development to prevent uncontrolled AI systems from destabilizing the internet.

By Monday, China's Ministry of Foreign Affairs fired back. Spokesperson Guo Jiakun dismissed Amodei's concerns as "fearmongering" and accused the U.S. of using AI governance as cover for containment, a Cold War playbook applied to the tech sector. The state-run Global Times newspaper said Amodei's essay was "packed with containment provisions targeting China."

This isn't an isolated spat. The exchange arrives as President Trump prepares for a late-September meeting with Xi Jinping where AI governance is expected to be on the agenda. It also follows a joint cybersecurity advisory from the FBI, National Security Agency, and Cybersecurity and Infrastructure Security Agency documenting what they described as "aggressive, malicious" efforts by Chinese AI developers to extract capabilities from advanced U.S. AI models like Anthropic's Claude and OpenAI's GPT.

What It Means for You

This isn't an abstract geopolitical drama. What the U.S. and China do about AI in the next 18 months will shape whether you have a job that still exists in a decade, whether the tools that diagnose your health are American or Chinese (and whether either are trustworthy), and whether the internet infrastructure you rely on remains functional or becomes a tool for large-scale harm.

Right now, you're caught between two bad options pretending they're the only ones available. Option one: play containment, restrict China's access to advanced chips and AI talent, and watch the global AI race accelerate anyway, China invests harder, America innovates faster out of competition anxiety, and safety takes a back seat everywhere. Option two: let competition run free, hope the best technology wins, and accept that the first country to build a genuinely dangerous AI system might not be constrained by democratic accountability.

Neither serves the common good. Both assume the game is zero-sum: that one country winning means the other loses. But AI governance isn't a war where there's a victor and a vanquished. It's an infrastructure problem. And infrastructure works best when the major players agree on standards.

The Bigger Picture

Here's what's actually true: Chinese companies have developed powerful AI models that are often cheaper to use than advanced models from U.S.-based companies, even as by some measures, they are just as capable. That's not a threat to national security. It's a sign that we're past the moment when the U.S. could monopolize advanced AI through restrictions alone.

The U.S. is not behind. It's not ahead in every metric either. It's in a competitive race with a country that has been investing heavily in AI for years and is now showing results. That's how competition works. The question isn't how to stop it. It's how to manage it so that both countries invest in safety instead of speed.

The real crisis is domestic. The U.S. has no federal AI law. Six companies control over 90 percent of frontier AI development. The European Union passed its AI Act in 2024. America hasn't started. And while the U.S. and China joust over who gets to build the most powerful system, neither has built the regulatory scaffolding to ensure it's safe when it does.

Chip export restrictions on China are a tactic dressed up as strategy. They slow China down short-term and accelerate the Chinese government's investment in domestic chip manufacturing long-term. They let American policymakers look tough while avoiding the harder work: building an AI regulatory framework that actually works, that the international community can live with, and that lets innovation happen inside real guardrails.

That's what Denmark does with labor market policy, it invests in retraining and support so automation doesn't destroy workers' lives. That's not passive accommodation. It's active management. The U.S. spends 0.1 percent of GDP on active labor market policies. Denmark spends 2 percent. One country treats automation as something that happens to people. The other treats it as something to be shaped by choice. The difference shows in people's lives.

Where This Goes

Trump and Xi will meet in late September. What happens in that conversation will ripple through every tech company in America and every AI researcher deciding where to work. The stakes are real. But so is the trap: framing this as containment versus surrender, when the actual choice is between outdated competitive logic and grown-up shared governance.

The Common Good Party's position is clear. China is both competitor and necessary partner. We reject the false choice between Cold War confrontation and naive accommodation. What we need instead is real regulatory leadership at home, a federal AI law with teeth, standards that the world can build on together, and a commitment that when the U.S. and China do cooperate on AI safety, neither side treats that as weakness.

That's harder than containment theater. It requires more sophisticated policy, more honest assessment of where we actually stand, and more willingness to lead by example instead of by restriction. It means building something instead of just blocking something. But it's the only approach that actually protects the common good, because the common good now includes not just Americans, but the fact that the world's two largest AI developers need to be able to talk about how not to break things.