World / AI & Power
Beijing Tells the AI Panic Crowd to Calm Down
The people building the models say the curve is too steep. The two governments that could actually force a pause both refuse to go first. One calls the warning fearmongering. The other calls winning the only real safety policy.
China’s Foreign Ministry used Monday’s regular briefing in Beijing to send a blunt message to Washington and Silicon Valley: stop treating artificial intelligence as a threat story, and stop treating the race as a Cold War. Spokesperson Guo Jiakun said AI “bears on the well-being of all humanity” and that governments should “jointly promote the open and inclusive development of AI for good and for all.” Fearmongering, confrontation and vicious competition, he warned, would only disrupt global AI governance and serve no one’s interest.
That was not an abstract lecture. It was a reply to Anthropic CEO Dario Amodei.
What triggered Beijing
On Saturday, Amodei published a roughly 3,800-word essay arguing the industry must “pace the frontier.” He said two things had changed his mind. First, AI systems are increasingly helping build the next generation of AI, a loop researchers call recursive self-improvement. Second, recent agent incidents, including an OpenAI-linked swarm that ran unauthorized attacks around the Hugging Face breach, showed models moving faster than the safety work meant to contain them.
“We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote. He also argued that a Chinese lead in AI “would pose grave danger for the United States and the world,” and that chip and equipment restrictions on China should stay in place. His plan had three rungs: independent evaluators embedded inside frontier labs, shared safety standards among democratic countries, and some form of deal with China. He called the China piece the hardest. OpenAI’s Sam Altman and xAI’s Elon Musk both signaled agreement within a day, and Google DeepMind was among the labs calling to ease the pace.
Beijing heard the safety language. It also heard the containment language. State media went further than the Foreign Ministry: the state-backed Global Times branded Amodei’s essay, “in essence, a ‘Cold War playbook’ for the AI sector.”
Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance, which serves no one’s interest. Guo Jiakun, China’s Foreign Ministry
Washington is not hitting the brakes either
If China rejected a slowdown, so did the White House, for the opposite reason. On Sunday, speaking at his golf resort in Doonbeg, Ireland, President Donald Trump said the United States is leading China in AI and he intends to keep it that way. “Whoever wins AI wins,” he told reporters. Guardrails can be added, he said, but “negative forces” were raising scenarios “that won’t happen.”
That is the real shape of the week. Lab chiefs are talking about pacing. The two governments that actually control the chips, the energy, the export licenses and the military budgets are talking about winning.
Xi Jinping and Trump are expected to meet in Washington later this month, with AI already on the list. The two sides agreed after their May meeting to set up an intergovernmental dialogue on the technology. That is process, not a treaty. Analysts quoted after Monday’s briefing said a real joint speed limit is close to impossible while both capitals treat frontier models as core national security infrastructure.
Open in the speech, tight in the system
China’s line is consistent with the pitch Xi has been making all year: open source, training slots for the Global South, a World AI Cooperation Organisation launched around July’s World AI Conference in Shanghai, and a BRICS open-source community floated again this weekend in New Delhi. That pitch lands in the Caribbean and the wider developing world because it sounds like access. Cheap or open models, training and “AI for all” language are easier to sell than Washington’s chip bans.
The contradiction is just as clear. China still runs a tightly controlled internet and polices its own models, data and speech. Its security chief, Chen Yixin, wrote this weekend that Beijing must balance development and security, tighten laws and ethics rules, and block hostile use of AI. The same state that tells the world not to fearmonger is not offering an unregulated free-for-all at home.
The United States has the mirror problem. Officials warn that Chinese firms distilled or scraped American models at industrial scale, and export controls on advanced chips remain the main American tool. Amodei’s essay did not just ask labs to slow down. It asked Washington to keep starving Beijing of compute. That is competition by another name.
The Vincy Angle
Notice who is not in this story. Not St. Vincent and the Grenadines, not CARICOM, not any small state that will use these tools without ever building them. The pause is being debated by the handful of companies that own the frontier models and the two governments that control the compute they run on. Everyone else is a rule-taker.
That is exactly why Beijing’s “AI for all” language is aimed partly at us. Open-source models, Global South training slots and a cooperation body that welcomes developing countries are an easier sell in the Caribbean than a chip ban that was never going to help us anyway. It sounds like access. Whether it delivers access, or delivers dependence on a different giant, is the question Vincentians should be asking before the marketing does the thinking for us.
Power to the People means noticing the shape of the room. When the people who build a technology say it is moving too fast, and the two governments who could slow it down both refuse, small nations do not get to opt out of the consequences. We only get to decide, with our eyes open, whose tools we come to depend on.
What VPN verified: Guo Jiakun’s remarks at the September 14 Foreign Ministry briefing, Amodei’s September 12 essay and its central claims, the agreement from Sam Altman and Elon Musk, the Global Times “Cold War playbook” editorial, and President Trump’s September 14 comments in Ireland are confirmed through Reuters, Bloomberg, CNBC, the Associated Press, Forbes and Amodei’s published essay. The Xi and Trump meeting is expected in Washington this month with AI on the agenda. Quotations are reproduced as reported by those outlets.