In the high-stakes race for AI supremacy, America cannot afford to be outmaneuvered. The threat posed by China’s accelerating AI capabilities demands a holistic, robust strategic response. The question is not whether America should beat China, but whether it should adopt a strategy of diffusion, abundance, and national strength, or outsource that strategy to a single company. Demanding that we trust Anthropic’s judgment, adopt their safety regime, and treat their corporate interests as synonymous with the national interest is as narrow as it is undemocratic. Anthropic’s latest policy paper, cloaked in the guise of national security analysis, does exactly this, revealing a self-serving agenda: manipulating genuine geopolitical risks to justify regulations that would cement the company’s own market dominance. The company’s core bet is that the United States must first establish a strong domestic AI safety regime, then, on that basis, negotiate against China. In this telling, America wins through frontier AI being governed through strict regulatory red tape. This isn’t just hypothetical; we have seen this in their own lobbying at the state and federal level for testing mandates, compute controls, reporting requirements, and institutionalized oversight. In their playbook, once the U.S. builds this regulatory thicket that only Anthropic can navigate, Anthropic…I mean, America, will be in a position to stand up to China. This Anthropic playbook frames the U.S.-China AI competition across four fronts — intelligence, domestic adoption, global distribution, and resilience. They declare that intelligence is the decisive factor. This framing is no neutral assessment. It is a calculated effort to elevate the very capabilities that Anthropic already claims to excel in: closed, frontier systems powered by massive compute clusters and proprietary safety processes. The company conveniently ignores that China is rapidly closing the gap with recent breakthroughs, or that half of the world’s AI researchers are in China. The underlying logic is clear: if intelligence is supreme, then the ones building the most advanced closed models become the globe’s indispensable champions, requiring protection and deferential regulation. Yet their own analysis undermines this premise. A few lines later, they admit, “intelligence alone is not sufficient.” China can overcome an intelligence deficit through faster, cheaper, and more globally distributed deployment of “good enough” AI. The problem is that most enterprises don’t need the absolute cutting edge; they need solutions that solve their problems at a price they can afford, with the flexibility to customize and deploy as needed. A model delivering 80% of frontier capability at 20% of the cost will capture far more market share than one offering marginal improvements at premium prices. This is the genuine Chinese threat, not surpassing America at the bleeding edge, but flooding global markets with “good enough” AI that becomes the default choice for businesses, developers, and governments worldwide. Anthropic’s dismissal of China’s open-weight models as currently lagging in adoption reveals a dangerous blindness to the dynamics of market disruption. Yes, American closed frontier models lead today. Enterprises are paying real money for our leading systems. But corporate leaders are already looking at their AI bills and asking hard questions. The value is real, but so is the cost. If an enterprise can get much of the same value from an open model at a fraction of the price, that option becomes attractive very quickly. The company’s treatment of open-source AI development reveals its true agenda. They warn that Chinese labs release dual-use capable models as open weights, allowing users to strip away safety guardrails. This concern has merit, but Anthropic’s implied solution is to restrict open-source AI development broadly. This would devastate American startups, universities, and small businesses while conveniently protecting their market position. Anthropic uses export controls as a rhetorical cudgel to justify a broader regulatory framework that would entrench its own position. The same pattern appears in Anthropic’s solutions. Close loopholes in export controls? Fine. Strengthen enforcement against chip smuggling, foreign data center workarounds, and unlawful diversion? Yes. Defend American models from distillation attacks? Absolutely. However, what American policymaker or serious U.S. lab is “allowing” China to steal capabilities? Export controls and anti-distillation defenses can be tools, not a full strategy. They may slow China in the short term but, as we have seen in the telecommunications industry, they alone cannot make the world dependent on American technology. These controls will not create global developer loyalty. They cannot ensure that enterprises, governments, and startups choose American AI over cheaper Chinese alternatives. In the long term, if we follow Anthropic’s prescription, this approach can backfire. Accelerating China’s drive for self-sufficiency, inspiring a parallel chip ecosystem, and pushing global markets toward non-American alternatives. The goal should be for American models and our computing capacity advantage to become the global standard. A policy that walls off markets without an affirmative diffusion strategy will enable our adversaries to fill the vacuum, and the U.S. loses influence over critical infrastructure. Maybe that is what Anthropic wants. The stakes in the AI race are too high for such transparent self-interest. Policymakers must look past the veneer of national security analysis and scrutinize whose interests are truly being served. The future of American innovation and global competitiveness hangs in the balance. Allowing one company’s corporate strategy to define the rules of the game would be a strategic catastrophe. *** Nathan Leamer is the executive director of Build American AI and a former policy advisor to FCC Chairman Ajit Pai.