Trump's AI Safety Plan: Slow China, Not America — SkimNews
Get the Geopolitics newsletter
Daily geopolitics — wars, elections, sanctions, the diplomatic moves that move markets. Free.
- Dario Amodei published "We Must Pace the Frontier" on September 12, urging leading labs to deliberately slow capability gains; Sam Altman said OpenAI would match Anthropic's commitment to host independent evaluators and Elon Musk replied "Dario is right."
- Donald Trump called AI catastrophe fears a hoax two days later, claiming China was the only party "happy" about AI backlash, while his 2024 voters preferred staying in the AI race (50%) over prioritizing safety (38%) per Politico polling.
- Chinese governance figures Xue Lan and Zeng Yi joined a Capitol Hill discussion convened by Senator Bernie Sanders this spring, calling for U.S.-China "safe zones" to build shared safety standards while cataloguing 94 classes of AI threats through Zeng's Beijing Institute of AI Safety and Governance.
- The NSA, FBI, and CISA accused six Chinese developers including DeepSeek and Alibaba of industrial-scale distillation of American frontier models, while the Trump administration simultaneously cleared roughly 10 Chinese companies to buy Nvidia H200 chips — which Chinese customs kept out.
- Treasury Secretary Scott Bessent proposed a standing U.S.-China AI dialogue with an incident notification mechanism on September 20 in New York, and on September 15 rejected the liability exemption labs are seeking before the House Financial Services Committee.
- China's Commerce Ministry threatened "resolute countermeasures" if Washington uses distillation accusations to suppress Chinese firms, and could resurrect rare earth export controls paused under the Busan truce; Beijing's Global Times called the slowdown push a "Cold War playbook."
Why it matters: The administration can claim it is acting on AI risk — 63% of Americans see a real chance AI could wipe out humanity, per Politico — while asking nothing of American labs. But if Beijing reads 'AI safety' as a synonym for containment, the standing incident channel Bessent proposed on September 20 goes quiet exactly when it is needed, and the modest deliverable Amodei suggested — a no-AI-for-biological-weapons pact — dies with it.
Ask SkimNews
