Beijing avviser Amodeis AI-bremse som «fryktspredning» — samtidig som Kina advarer mot AI-risiko hjemme

Beijing avviser Amodeis AI-bremse som «fryktspredning» — samtidig som Kina advarer mot AI-risiko hjemme

Da Anthropic-sjefen Dario Amodei i helgen oppfordret AI-bransjen til å senke farten, kom svarene raskt — og de spriker ikke bare mellom land, men innad i Kinas egen politikk. Ukenes AI-storm lander dessuten få dager før Trump–Xi-møtet 24. september.

Hva Amodei foreslo

Amodei publiserte essayet «We Must Pace the Frontier» lørdag 12. september 2026. Kjernebudskapet er at AI-selskapene bevisst må sette ned tempoet slik at risikohåndteringen rekker å følge med:

«We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.»

Han begrunner det slik: full risikohåndtering krever mer forsiktighet — ikke bare investering i risikoforebygging, men også å regulere hastigheten på kapabilitetsutviklingen slik at forebyggingen får tid til å hente igjen (darioamodei.com).

Ifølge NBC News advarte Amodei også om at et kinesisk overtak i AI «would pose grave danger for the United States and the world», og oppfordret USA til å begrense salg av nøkkelteknologi til Kina. Dette gjengis via NBCs sammendrag av essayet og er ikke direkte sitert fra primærkilden.

Beijings tosporede svar

Utenriksministeriet tok til motmæle allerede mandag 14. september. Talsmann Guo Jiakun sa til journalister:

«Fearmongering, confrontation and malicious competition will only disrupt the process of global AI governance and serve no one's interests.» (NBC News)

Allerede søndag hadde statskontrollerte Global Times gått hardere ut. Redaksjonen kalte ifølge NBC forslaget om bransjeomfattende nedfart for «packed with containment provisions targeting China» og «in essence, a 'Cold War playbook' for the AI sector». Redaksjonens fulle tekst foreligger ikke — bare NBCs gjengivelse.

Men samtidig dukker Kinas eget risikofokus opp på andre sider av det offisielle ordskiftet. Statssikkerhetsminister Chen Yixin sa søndag at Kina må «accelerate the establishment of a system for preventing and controlling AI security risks», og skrev i det statlige magasinet China Cyberspace at Kina må «engage extensively in international cooperation» — mens han samtidig omtalte AI som «a new track for strategic rivalry» (NBC News).

Spenningen mellom utenriksretorikken og den interne politikken er ikke bare et observatørinntrykk. Ifølge Reuters via The Straits Times inkluderte en kinesisk AI-sikkerhetsramme utgitt i september 2024 under Cyberspace Administration of China for første gang et eksplisitt fremtidsscenario om at AI kan komme ut av menneskelig kontroll. Beijing fornekter altså ikke selve risikoen Amodei advarer mot — det avvisningen rammer, er om nedfart i vesten tjener andres interesser enn Kinas.

Trumps motangrep

President Trump har avvist advarslene helt, både til reportere og på Truth Social. Til reportere søndag sa han: «We're leading China in AI. We're the most sophisticated country in the world, and frankly I want to keep it that way because whoever wins AI wins» (NBC News).

På Truth Social gikk han til angrep på Amodei ved navn. Ifølge New York Post skrev Trump at «There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!» — og kalte ifølge avisen risikoen for at AI kan ødelegge menneskeheten en «HOAX».

Blant bransjeledere var tonen en annen: OpenAIs Sam Altman støttet Amodeis forslag om å gi uavhengige evaluere ansattekvivalerende tilgang, og Elon Musk sa at Anthropic-sjefen hadde rett (Bloomberg via Yahoo Finance).

Bakgrunnen som spisser konflikten

Essayet kom få dager etter at amerikanske etater publiserte en advarsel om at toppkinesiske AI-selskaper systematisk har kopiert avanserte amerikanske modeller — en beskyldning Beijing har avvist (NBC News). Kopieringsadvarselen og Amodeis oppfordring om teknologirestriksjoner gjør at Kina i praksis svarer på et pakkeangrep av kritikk, ikke bare ett essay.

Veien videre: Trump–Xi 24. september

Spørsmålet er hva som skjer når Trump møter Xi Jinping 24. september. Ifølge NBCs dekning skal AI være på agendaen, men det er usikkert hvor konkret nedfart, sikkerhetsrammer eller teknologi­restriksjoner blir berørt i møtet. Det dokumenterte grunnlaget så langt er dette: Amodei vil bremse, Trump vil vinne løpet, og Beijing avviser «fryktspredning» utad — mens Kinas egne sikkerhetsrammer siden 2024 har forberedt seg på nettopp det scenariet Amodei mener krever nedfart.

---

Beijing dismisses Amodei's AI slowdown as 'fearmongering' — while China warns of AI risks at home

When Anthropic CEO Dario Amodei called on the AI industry to slow down over the weekend, the responses came quickly — and they diverge not only between countries but within China's own policy. The week's AI storm also lands just days before the Trump–Xi meeting on September 24.

What Amodei proposed

Amodei published the essay "We Must Pace the Frontier" on Saturday, September 12, 2026. Its core message is that AI companies must deliberately reduce their pace so that risk management can keep up:

"We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain."

His reasoning: full risk management requires more caution — not only investing in risk prevention, but also regulating the speed of capability development so that prevention has time to catch up (darioamodei.com).

According to NBC News, Amodei also warned that a Chinese lead in AI "would pose grave danger for the United States and the world," and urged the U.S. to restrict sales of key technology to China. This is reported via NBC's summary of the essay and is not directly quoted from the primary source.

Beijing's two-track response

The Foreign Ministry pushed back as early as Monday, September 14. Spokesperson Guo Jiakun told journalists:

"Fearmongering, confrontation and malicious competition will only disrupt the process of global AI governance and serve no one's interests." (NBC News)

On Sunday, the state-controlled Global Times had already gone harder. According to NBC, the editorial called the proposal for an industry-wide slowdown "packed with containment provisions targeting China" and "in essence, a 'Cold War playbook' for the AI sector." The editors' full text is not available — only NBC's rendering.

Yet at the same time, China's own risk focus appears on other sides of the official discourse. State Security Minister Chen Yixin said on Sunday that China must "accelerate the establishment of a system for preventing and controlling AI security risks," and wrote in the state magazine China Cyberspace that China must "engage extensively in international cooperation" — while also describing AI as "a new track for strategic rivalry" (NBC News).

The tension between the foreign-policy rhetoric and domestic policy is not merely an outside observer's impression. According to Reuters via The Straits Times, a Chinese AI safety framework published in September 2024 under the Cyberspace Administration of China for the first time included an explicit future scenario in which AI could escape human control. In other words: Beijing is not denying the risk Amodei warns about — what the rejection targets is whether a slowdown in the West serves anyone's interests other than China's.

Trump's counterattack

President Trump has rejected the warnings outright, both to reporters and on Truth Social. To reporters on Sunday, he said: "We're leading China in AI. We're the most sophisticated country in the world, and frankly I want to keep it that way because whoever wins AI wins" (NBC News).

On Truth Social, he attacked Amodei by name. According to the New York Post, Trump wrote that "There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!" — and, according to the paper, called the risk that AI could destroy humanity a "HOAX."

Among industry leaders, the tone was different: OpenAI's Sam Altman backed Amodei's proposal to give independent evaluators employee-equivalent access, and Elon Musk said the Anthropic CEO was right (Bloomberg via Yahoo Finance).

The backdrop that sharpens the conflict

The essay came days after U.S. agencies published a warning that top Chinese AI companies have systematically copied advanced American models — an allegation Beijing has rejected (NBC News). The copying warning and Amodei's call for technology restrictions mean that China is effectively responding to a package of criticism, not just one essay.

The road ahead: Trump–Xi on September 24

The question is what happens when Trump meets Xi Jinping on September 24. According to NBC's coverage, AI is expected to be on the agenda, but it is uncertain how concretely pacing, safety frameworks, or technology restrictions will come up in the meeting. The documented basis so far is this: Amodei wants to slow down, Trump wants to win the race, and Beijing rejects "fearmongering" externally — while China's own safety frameworks since 2024 have been preparing for precisely the scenario Amodei argues requires a slowdown.