Amodei's AI slowdown pitch rebuffed by Guo Jiakun

For years, many of Silicon Valley's most powerful figures framed AI as an arms race with China whose loss would be disastrous. Over the past week, some of those same executives shifted emphasis, arguing that AI could become dangerous enough that the two countries need to work together to slow the race. Amodei's blog post is the clearest statement of that shift, but it came bundled with a condition: the slowdown strategy should ensure the US and other democracies maintain "as large as possible" a lead over autocracies like China.

That pairing explains much of the response. A technology hedge fund founder, Kevin Xu, describes a gulf between an Anthropic-led vision that would restrict China first to maximize leverage before talking and a China that would only talk from a position of equals. Alvin Wang Graylin of the Asia Society Policy Institute's Center for China Analysis says the conditional framing makes the proposal feel inauthentic to Chinese readers. Meanwhile, Beijing's own anxieties center on domestic stability and party control rather than the doomsday scenarios that preoccupy US frontier labs, and China's AI industry has broadly focused on releasing lower-cost, open-weight models.

Even so, cooperation is not foreclosed. Preliminary negotiations are reportedly underway ahead of Xi's planned Washington visit, and both governments share worries about cybersecurity, misuse of powerful models, and capable systems behaving unexpectedly. What the talks hinge on is whether each side can verify the other's commitments — and whether the rhetoric about weakening China erodes the trust those talks depend on.