China bogeyman looms lairge ower American firms’ AI day-o-reckonin scenario

When reporters speired at Donald Trump this week gin he backit calls tae slow doun the development o artificial intelligence oot o growin fears for cybersecurity, public safety and the fate o humanity, he said nae. His argument: China.

“We’re leadin China in AI,” Trump said. “We’re the maist sophisticated kintra in the warld, and frankly, I want tae keep it that gate, acause whaever wins AI, wins.”

Trump wis echoin a message the AI industry has put oot for years. The US is in a het technological race wi China, companies includin Anthropic and OpenAI hae argued, and ony regulation that wad cede the contest risks incalculable economic and political hairm.

Even as Dario Amodei, Anthropic’s CEO, led an industry-wide call this week tae “pace the frontier”, he declared that the US cud only slow doun tae the point whaur China cudna catch up.

“Gin we greatly restrain oor AI capabilities in the belief that China will dae the same, and syne China defects, AI cud be sae pouerful that sic a defection cud lead tae their geopolitical dominance,” Amodei warned.

The AI industry’s concerns aboot China may come wi self-servin economic incentives, but they also reflect a strain o hawkish foreign policy toward Beijing that has become entrenched in tech.

In recent years, pairts o Silicon Valley hae positioned theirsels as the vanguard in a new cauld war wi China, arguin fiercely that the industry’s technological advancements are essential tae baith US hegemony and the liberal warld order. This view has taken hauld across tech’s different ideological camps, creatin strange bedfellows o AI safety advocates, techno-militarists and neoconservatives, a’ o them fearful o Beijing.

AI companies hae hired former tap US government staffers frae baith the Biden and Trump administrations tae shape public strategy and analyze geopolitical risk, settin up in-hoose policy teams. Influential tech leaders hae meanwhile written essays and policy documents callin for efforts tae deter and degrade China’s AI capabilities, sometimes advocatin for threatenin airmed conflict gin necessary, in an attempt tae influence foreign policy.

As public calls for reinin in the AI industry intensify and the risks associated wi the technology increase, various industry forces are tryin tae ensure they are the anes wi the answer tae the questin “whit aboot China?”.

Anthropic and OpenAI didna respond tae requests for comment for this airticle.

Few hae been mair vocal aboot the threat o Chinese state-controlled artificial intelligence than Amodei. The Anthropic CEO has argued for years that ane o the reasons he’s sae animated tae advance AI is tae ensure that the technology falls intae the hands o Democratic governments and no oppressive regimes sic as China.

Amodei’s position has made him ane o Silicon Valley’s maist vocal critics o China’s tech industry, separatin him frae fallow AI leaders sic as OpenAI CEO Sam Altman or Meta founder Mark Zuckerberg, wha frame the competition mair in corporate or economic terms; and folk sic as Jensen Huang, Nvidia’s CEO, wha hae encouraged the use o Chinese AI models. Demis Hassabis, the leader o Google Deepmind, has made few public remarks on China in comparison wi his peers.

“Sam Altman disna hae especially strang views when it comes tae China. Anthropic is a totally different story,” said Kyle Chan, a fallow at the Brookings Institution. “Amodei is really, really focused on a particular view o the risk frae AI and whit he frames as basically a struggle atween a warld dominated by authoritarian AI versus democratic AI”.

Amodei’s argument, which he has laid oot ower a series o essays, is that China acquirin superpouered AI is ane o the maist severe threats posed by the technology. He believes humanity is approachin the creation o an ultra-advanced AI, and that sic a tool wad gie an authoritarian state the ability tae create a “swarm o millions or billions o fully automated airmed drones” that wad be “capable o baith defeatin ony military in the warld and suppressin dissent within a kintra by followin aroond every ceetizen”.

“There are a nummer o possible arguments against the severity o these threats, and I wish I believed them, acause AI-enabled authoritarianism terrifies me,” Amodei wrote in Januar in an essay cried The Adolescence o Technology.

Amodei has lang asserted that in order tae stap China’s advancement, the US and its allies shud take aggressive action tae limit Chinese AI companies’ growth – maist notably blockin the sale o advanced semiconductor chips and manufacturin equipment necessary for improvin AI. (Huang o Nvidia has bristled at this position, callin export controls a failure and urgin Trump tae ease restrictions.)

But Amodei’s position wis perhaps maist on display last weekend, in his response tae warnins frae a young Anthropic researcher wha resigned, predictin human extinction by the end o the decade at the digital hands o AI. The apocalyptic string o tweets by Jacob Coxon didna mention China. But the specter across the Pacific looms sae lairge in the Anthropic CEO’s mind that, when outlinin hou tae prevent human extinction, beatin Beijing wis paramount.

Amodei has built a team that shares his views. He for example hired Tarun Chhabra, former Biden-era national security council staffer, as Anthropic’s heid o national security policy. Chhabra has campaigned against the sale o chips tae China and has cried for a full-scale government effort tae curb the Chinese Communist pairty’s AI advancements.

“The CCP disna seek complementarity or coexistence or comparative advantage wi the United States or oor allies. Instead, the CCP seeks run-the-table dominance and control o the maist critical technologies and supply chains,” Chhabra tauld a Senate hearin last year, addin that he believed the Biden administration didna take decisive or rapid enough action tae maintain the US lead against China.

Amodei’s concern ower export controls has also placed him in the company o ither influential China hawks. He wrote an op-ed for the Wall Street Journal last year wi Matt Pottinger, the airchitect o Trump’s first-term China policy, urgin the administration tae clamp doun on chip exports. Pottinger, a visitin fallow at the conservative Hoover Institution and author o a buik on hou tae defend Taiwan against a Chinese attack, argued in 2024 that the US and China were in the “foothills o a great-pouer het war”.

Anthropic and Amodei’s stance on China puts the company in strange tension wi its ain domestic policy. The company has spent millions lobbyin for increased government oversight and safety measures. At the same time as Anthropic advocates for mair safeguards, it has made clear that it will aggressively oppose ony policy it perceives as giein China the chance tae overtake the US.

“There’s somethin that’s really important that everyone hears frae Anthropic,” Sarah Heck, the company’s heid o public policy, said at a Politico event on Wednesday. “The United States needs tae stay in the lead on AI and ye canna dae safety frae second place.”

Understandin the beliefs o AI companies and CEOs sic as Amodei is particularly important, experts say, acause the Trump administration’s laissez faire policy towards the AI industry has meant that private companies are the primary actors in shapin public policy.

“It wad be akin tae haein the entire pharmaceutical industry be just completely unregulated,” Chan said. “It’s kind o amazin and ultimately boils doun tae ane thing, which is Trump has been very reluctant tae want tae regulate acause he disna want tae slow doun the AI boom and the effect that it’s haein on the economy.”

Anthropic isna the only company tae hire and align wi former tap policy staffers, wi OpenAI hirin the author o Trump’s AI action plan Dean Ball as “heid o strategic futures” earlier this year. While Ball has criticized Amodei’s framin o the competition wi China, he has expressed a similar pessimism on US-China cooperation.

“There wis a time when it wis possible tae believe we cud each pursue oor strengths, enrich oor respective economies, and grow thegither. Alas, sic harmony nou appears impossible,” Ball wrote o China last year. “We are locked intae a structural conflict.”

Although Amodei is the AI industry’s maist prominent voice on China, fears o Beijing gainin advanced AI capabilities and usin them for global domination run through Silicon Valley.

Several o the tech figures focused on China hae connections, like Amodei hissel, tae the effective altruism movement – a quasi-utilitarian idea that society shud focus on projects that maximize the guid and reduce the maist hairm for humanity. The movement tends tae overlap wi AI safety advocates wha worry aboot the lang-term, existential risks o the technology.

Leopold Aschenbrenner, a 25-year-auld former OpenAI researcher, founder o an influential AI hedge fund, and husband tae Amodei’s chief o staff, cried for a government-led effort tae reach super-intelligent AI before China in a 165-page paper.

“Super-intelligence will gie a decisive economic and military advantage. China isna at a’ oot o the game yet. In the race tae AGI, the free warld’s very survival will be at stake,” Aschenbrenner wrote in 2024.

The position o Aschenbrenner, wha made headlines earlier this year when his hedge fund lost $35bn in less than a month, wis echoed in Washington. When the typically hawkish US-China economic and security review commission issued its annual report tae Congress a few months after Aschenbrenner’s manifesto, the congressional advisory committee’s nummer ane suggestion wis tae “establish and fund a Manhattan Project-like program dedicated tae racin tae and acquirin an Artificial General Intelligence”.

Grand strategy proposals frae the AI industry on China are no solely contained within the effective altruist movement. In 2025, Google co-founder Eric Schmidt, Alexander Wang, Scale AI founder (nou heid o AI at Meta), and Dan Hendrycks, Center for AI Safety director, wrote a wide-rangin strategic deterrence doctrine cried Mutual Assured AI Malfunction (Maim). Their paper applies the cauld war framework o Mutually Assured Destruction (Mad) wi nuclear weapons and applies it tae AI, proposin that the US and China cud achieve geopolitical stability through threatenin espionage, cyberattacks and potential military strikes against each ither gin either state pursues a monopoly on AI.

“A nation wi sole possession o super-intelligence might be as overwhelmint as the Conquistadors were tae the Aztecs,” the paper warns.

Some leadin AI thinkers hae argued for extreme measures tae deal wi AI advancement frae nations sic as China. Eliezer Yudkowsky, ane o the earliest and loudest voices o the AI safety movement, has argued that the US shud be willin tae carry oot airstrikes against foreign data centers and risk nuclear war in order tae make rival states comply wi a global AI moratorium.

“Make it explicit in international diplomacy that preventin AI extinction scenarios is considered a priority abuin preventin a full nuclear exchange, and that allied nuclear kintras are willin tae run some risk o nuclear exchange gin that’s whit it takes tae reduce the risk o lairge AI trainin runs,” Yudkowsky wrote in a 2023 Time Magazine essay.

Yudkowsky foondit the Machine Research Intelligence Institute, which has received mair than $10m in fundin ower the past decade frae Open Philanthropy, a non-profit associated wi the effective altruism movement. Open Philanthropy’s cofounder and former CEO Holden Karnovsky is mairried tae Anthropic co-founder Daniela Amodei.

Ither China hawks in the tech and AI industry tend tae hae ties tae the military industrial complex and an economic incentive in the competition, sic as Anduril founder Palmer Luckey and Palantir CEO Alex Karp. Luckey has claimed that his defense tech firm has an internal policy that ony o its projects need tae be built wi the assumption China will launch hostilities against Taiwan in 2027.

Karp has meanwhile framed the AI competition wi China as a fight for geopolitical supremacy and suggested Americans shud accept tradeoffs tae their civil liberties tae win.

“We are goin tae be the dominant player or China is goin tae be the dominant player. And there will be very different rules dependin on wha wins,” Karp tauld Axios last year. “When folk are worried aboot surveillance, o course there are huge dangers there, but it’s like, ye will hae far fewer richts gin America isna in the lead.”

While Amodei and ither US AI leaders frame competition wi China as a race tae reach a super-pouerful artificial intelligence first and worry aboot AI breakin free o human control, experts say that maist Chinese firms and their government dinna view the technology the same gate. Instead, loss o control debates are relatively new in China, and the government’s primary concern is hou AI cud lead tae domestic social unrest.

As China and the US are set tae hauld talks this month on AI cooperation, experts say Beijing is primarily approachin the talks frae a place o distrust that the negotiations are merely a gate tae suppress its growth under the guise o safety. Amodei’s essay has fuelled that concern.

“They’re very muckle payin attention tae Amodei’s essay and they’re also conflatin his essay wi potential US approach in the upcomin AI safety talks,” said Kenton Thibaut, senior resident China fellow at the Atlantic Council’s Digital Forensic Research Lab. Trump is set tae meet wi Xi Jinping in less than a week in Washington, wi AI high on the agenda.

China’s state-backit media properties hae aggressively criticized Amodei, accusin him o runnin a “cauld war playbook” against China through his support for an AI slowdown. Experts say Anthropic’s dual calls for baith strict export controls on China and safety agreements hae also had the effect o combinin the discourse aroond twa arguably separate subjects: AI safety and economic competition aroond AI.

“Anthropic has played quite a strang role here in linkin those issues thegither in a gate that they maybe didna need tae be linked,” Brookings’ Chan said. “Tae put it really bluntly, it made it muckle harder tae get the Chinese tae want tae talk aboot these things and no see it purely as a veiled gate tae slow them doun.”

While the Trump administration negotiates wi China, the growin domestic backlash tae the AI industry in the US may force lawmakers tae take action regardless o hou it affects global pouer dynamics. Nationwide opposition tae datacenters and recent cybersecurity incidents sic as OpenAI’s models hackin intae ootside organizations hae become midterm election topics, addin incentive for lawmakers tae address the immediate hairms related tae AI.

“Ye can gang ahead and wield the China airms race argument, but at some point this industry will also hae tae deal wi the consequences o fuelin a race tae the bottom,” said Amba Kak, co-executive director at the AI Now Institute and former senior adviser on AI for the Federal Trade Commission.

Although experts and diplomats hae expressed little hope that upcomin US-China talks will lead tae major agreements, it is also possible that the increasingly real risks o AI bein used for autonomous weapons, cyberattacks and ither destabilizin technologies may force kintras tae shift the AI race’s course even gin they dinna stap buildin it.

“We dinna win gin we develop the killer robots first. Naebody wants killer robots,” Thibaut said. “There shud be pouerful incentives on baith sides tae stap this.”