China's security establishment responded to US tech leaders' calls for an AI slowdown with a coordinated but internally contradictory message this weekend: the Foreign Ministry dismissed the pacing push as "fearmongering," while the country's top intelligence official published one of Beijing's most detailed warnings yet about the very same risks the American CEOs cited. ANALYSIS The simultaneity exposes a dual-track posture: China will resist any externally imposed brake on AI development while quietly building its own domestic risk-prevention apparatus.
Why it matters
The US and China are the two dominant forces in frontier AI development7. How each frames AI risk, and whether either accepts constraints, will shape the trajectory of global governance. A planned meeting between President Donald Trump and Chinese leader Xi Jinping on September 24 is expected to include AI governance among the topics discussed5. ◆ The signals Beijing sent this week amount to a pre-negotiation position: safety is real, but only on our terms.
The big picture
On Saturday, Anthropic co-founder Dario Amodei argued in a blog post that AI capabilities are outpacing safeguards and proposed coordination between companies, the US, and other democracies9. He warned that a "Chinese lead in AI would pose grave danger for the United States and the world" and called for continuing restrictions on sales of cutting-edge AI chips and chipmaking equipment to China. OpenAI CEO Sam Altman and Elon Musk joined the chorus demanding industry-wide pacing1,4.
Beijing's Foreign Ministry fired back on Monday. "Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance which serves no one's interest," spokesperson Guo Jiakun said at a regular press conference. The CEOs' warning caused AI stocks to move lower and drew a rebuke from President Trump, who said the US needed to maintain its edge over China.
Yet one day before Guo's dismissal, State Security Minister Chen Yixin published an article in China Cyberspace magazine that catalogued AI dangers in striking detail8,11. Abusive use of AI by adversaries could "directly threaten China's political security, institutional security and ideological security," Chen wrote. He cited AI's increasing ability to discover flaws in systems and create malware or software to exploit those weaknesses. He named specific US-developed models: Anthropic's Mythos and OpenAI's GPT-5.5-Cyber could lower the technical threshold and costs of carrying out cyberattacks, he warned6,10.
Chen's threat inventory went further. He warned that "hostile forces" could wage "public-opinion wars" and "cognitive warfare" using deepfakes, AI-generated text and images, and social media bots to fabricate political rumours, spread harmful information, and incite social divisions. Foreign intelligence services, he wrote, are actively using intelligent web crawlers, data-mining technologies, and profiling to collect sensitive information, including government data, commercial secrets, and personal information. He called for the creation of reliable risk-prevention mechanisms and stronger global governance of AI.
ANALYSIS Chen's article and Guo's press conference were not contradictory by accident. They serve distinct audiences with a consistent strategic logic. The Foreign Ministry message is aimed outward: any US-led framework that slows development is treated as a competitive ploy, not a safety measure. Chen's article is aimed inward and at the security bureaucracy: the risks are real, the preparation must accelerate, but the threat originates from foreign models and foreign intelligence services, not from China's own AI programs.
Chen also warned that countries with an advantage in AI may invoke national security to impose technology controls, monopolize industry standards, and build closed-source ecosystems. ◆ This framing repackages Amodei's call for export controls as evidence of the very threat Chen is warning about, turning the American safety argument into a data point for Chinese strategic anxiety.
Chinese AI developers have promoted open-weight models partly on the grounds that cybersecurity teams can inspect, modify, and deploy them for defensive work. ◆ Chen's emphasis on the dangers of closed-source US models reinforces this domestic preference while casting American firms' proprietary approach as a security liability for China.
Chen's article made no mention of the calls by Anthropic and OpenAI to slow down the pursuit of cutting-edge models, and there is no sign the timing was coordinated with the US pacing debate. ◆ The parallel timing nonetheless creates a negotiating dynamic: Beijing can point to its own safety preparations as evidence of responsible governance while rejecting any external constraint.
What's next
The Trump-Xi meeting planned for September 24 is expected to put AI governance on the bilateral agenda. Chen Yixin called AI "a new arena for strategic rivalry among major powers". At the BRICS summit in India over the weekend, Xi Jinping proposed an AI-powered initiative to drive new industrialization and deepen supply chain cooperation in the emerging economies bloc. ◆ Beijing is building its case on two fronts simultaneously: bilateral talks with Washington and multilateral forums where it can set norms without conceding to US-led frameworks. The September 24 meeting will test whether either side's safety rhetoric translates into binding commitments or remains, as Guo put it, a matter of "fearmongering."