New York Times ChineseAug 17, 2026, 11:37 UTC+8

AI Arms Race: China Gains Fast as US Policy Wavers

人工智能军备竞赛：中国快速进步，美国政策摇摆

The Pentagon banned Anthropic from military systems over CEO Dario Amodei's refusal to drop restrictions on autonomous weapons and domestic surveillance, then walked it back within a month. The NSA kept access to the Mythos model for offensive cyber tests, calling a halt 'unilateral disarmament.' China may trail the US by only six months in frontier models and has shown more openness to AI arms control talks than it ever did on nuclear issues. The article does not detail any negotiation framework or timeline.

Why it matters: NYT exclusive on the Pentagon's ban-and-reversal dance with Anthropic, with named officials, a public secretary-CEO clash, and a concrete NSA dependency on Mythos. Hits the hardest nerve in AI safety and militarization. Minor ding: the 'China rapid progress' angle is barely de...

Read the original ↗Export Markdown

Take

This piece is worth reading because it lays out the messiness of the AI arms race in concrete terms. The Pentagon sent a letter in mid-July ordering military contractors to strip all Anthropic products from weapons and control systems by September 1. The trigger: CEO Dario Amodei demanded the military promise not to use its models for domestic surveillance or fully autonomous weapons. Defense Secretary Pete Hegseth publicly clashed with him over it. Then, less than a month later, the Pentagon quietly walked the ban back. The NSA had kept access to Mythos the whole time for offensive cyber tests, and one recently departed senior official called a halt 'unilateral disarmament.'

The other thread is China's pace. US estimates cited in the article put China roughly six months behind on frontier models, and Chinese officials have shown more openness to AI arms control talks than they ever did on nuclear issues. But the article doesn't detail any negotiation framework or timeline, so I'd treat that as a posture signal, not a policy shift.

I'd read this as two things colliding: Anthropic's usage restrictions got steamrolled by operational necessity, and even people inside the US government think exempting open models from security review makes no sense.

Source: New York Times Chinese · cn.nytimes.com
Published: Aug 17, 2026, 11:37 UTC+8 · Aug 17, 2026