The Anthropic China AI race argument has become so central to Washington’s technology debate that it is now doing something the AI industry could only have dreamed of a few years ago: substituting for regulation. That is a problem worth examining clearly.
When Dario Amodei published his essay ‘The Adolescence of Technology’ earlier this year, it did several things at once. It laid out a genuine, if contested, case for the existential risks of advanced AI. It warned that Anthropic’s own measurements suggest large language models may already be doubling or tripling the likelihood of success in bioweapons acquisition, a finding alarming enough that it shaped the release conditions for subsequent Claude models. And it declared, as a matter of priority, that beating Beijing trumps almost everything else, including, apparently, preventing human extinction.
That last move is the one worth dwelling on.
When Safety and Strategy Become the Same Argument
Amodei is not cynically waving a flag. His concern for democratic AI versus authoritarian AI is, by every account, sincere. But sincerity and strategic convenience are not mutually exclusive, and the China hawk position he has staked out carries enormous advantages for Anthropic as a company lobbying against regulation.
The logic runs like this: slow AI development for safety reasons and you hand Beijing an opening. Therefore, safety and speed are compatible only up to the point where China cannot catch up. Anthropic’s head of public policy, Sarah Heck, put it plainly at a Politico event: ‘The United States needs to stay in the lead on AI and you cannot do safety from second place.’
As a rhetorical move, it is close to impregnable. Any regulator who presses too hard can be painted as a de facto ally of the Chinese Communist Party. Kyle Chan, a fellow at the Brookings Institution, frames it precisely: ‘Amodei is really, really focused on a particular view of the risk from AI and what he frames as basically a struggle between a world dominated by authoritarian AI versus democratic AI.’
The argument has also had an unintended diplomatic cost. Chan’s assessment is direct: ‘Anthropic has played quite a strong role here in linking those issues together in a way that they maybe didn’t need to be linked. To put it really bluntly, it made it much harder to get the Chinese to want to talk about these things and not see it purely as a veiled way to slow them down.’
The Anthropic China AI Race and Its Broader Silicon Valley Coalition
Amodei has built an internal team designed to press this agenda at the highest levels of government. Tarun Chhabra, Anthropic’s head of national security policy, previously served as Deputy Assistant to the President and Coordinator for Technology and National Security on the National Security Council from 2021 to 2025. In that role, according to the Senate Foreign Relations Committee, Chhabra led the development of US policy on export controls, industrial strategy, and investment controls. He testified on 2 December 2025 before the Senate Foreign Relations Subcommittee on East Asia that ‘the CCP does not seek complementarity or coexistence or comparative advantage with the United States or our allies. Instead, the CCP seeks run-the-table dominance.’
That is a maximalist framing, and it is now Anthropic’s official posture.
The rest of Silicon Valley is not far behind. Leopold Aschenbrenner, whose 165-page manifesto called for a Manhattan Project-style race to artificial general intelligence, captured the mood. Google’s Eric Schmidt, Scale AI’s Alexander Wang, and the Center for AI Safety’s Dan Hendrycks co-authored a strategic doctrine called Mutual Assured AI Malfunction, applying the cold war logic of mutually assured destruction to AI, including the threat of military strikes against data centres.
Some have gone further. Eliezer Yudkowsky has argued openly for airstrikes against foreign data centres and for making it explicit in international diplomacy that preventing AI extinction scenarios ranks above preventing a full nuclear exchange.
Meanwhile, Amodei’s essay is being read in Beijing. Kenton Thibaut of the Atlantic Council’s Digital Forensic Research Lab notes that Chinese state actors are ‘conflating his essay with potential US approach in the upcoming AI safety talks,’ treating it as evidence that safety dialogue is cover for economic containment. China’s state media has accused Amodei of running a ‘cold war playbook.’
The irony is sharp. Anthropic’s own security report, flagged by House Homeland Security Republicans, assessed with high confidence that a Chinese state-sponsored actor used Claude tools to execute an autonomous cyberattack against a US company with minimal human involvement. The threat from Beijing is real. The question is whether framing every policy debate through that lens serves the public interest, or primarily Anthropic’s.
Amodei has also warned publicly that AI could displace half of all entry-level white-collar jobs within one to five years. That is a domestic crisis of the first order, and it sits uneasily alongside a corporate posture that treats any check on AI development as a gift to China. Amba Kak of the AI Now Institute offers the sharpest counter: ‘You can go ahead and wield the China arms race argument, but at some point this industry will also have to deal with the consequences of fuelling a race to the bottom.’
Trump meets Xi Jinping in Washington within days, with AI at the top of the agenda. If that meeting produces nothing, the pressure on Congress to act domestically will grow regardless of what happens across the Pacific.
