From Fortune · Emily Forlini · · 2m
Why Trump and Xi may skip an AI pause
OpenAI and Anthropic want a U.S.-China AI slowdown, but both governments see speed as power.
In brief
OpenAI and Anthropic want a U.S.-China AI slowdown, but both governments see speed as power. A U.S.-China AI safety deal is more likely to start with narrow talks on weapons, misuse, and testing than with any broad slowdown. Originally reported by Fortune.
Safety fears spike
U.S. anxiety over AI risk surged after a viral post warned the technology could end humanity by decade's end. OpenAI and Anthropic renewed calls for Washington and Beijing to slow the race.
“the technology could kill all of humanity by the end of the decade”
Altman floats a prize
Sam Altman told Fortune that Trump and Xi could win the Nobel Peace Prize if they struck an AI safety deal. The appeal has not moved either leader toward a slowdown.
“could win the Nobel Peace Prize”
Trump rejects limits
Trump said the U.S. needs a strong president, not AI guardrails. China dismissed the U.S. debate as fear-mongering that could block sound global governance.
“Whoever wins AI, wins!”
China sees containment
In Beijing, slowdown calls look like another U.S. attempt to protect its lead, much like chip export controls. Xi views AI as a rare chance to shift global tech power.
“China does not need U.S. permission”
Talks may start small
Experts expect AI safety to come up when Trump and Xi meet on Sept. 24. But a broad deal is far off because the countries lack shared rules for testing and oversight.
“a long way”
Weapons are the opening
The likelier agenda is narrower, including a pledge not to weaponize AI and principles to stop non-state actors from using models against finance or infrastructure.
“neither country wants to see”
Amodei draws backlash
Dario Amodei's call for an AI speed limit has split opinion. Some U.S. observers see a risk framework, while critics in China see a bid to preserve U.S. market power.
“speed limit”
Trump sides with speed
Trump dismissed Amodei's warning during a call with Nvidia CEO Jensen Huang. He also called fears of AI taking over a hoax and said progress would not stop.
“won’t stop our progress”
Open versus closed
DeepSeek engineer Shengyu Liu argued frontier AI should be open and cheap. That reflects a wider split, with U.S. labs favoring closed models and China pushing open-weight systems.
“available to everyone in an open and inexpensive form”
China is less alarmed
The U.S. debate about saving humanity has not taken hold with most Chinese citizens. Many see AI as useful, profitable, and tied to national progress.
“very positive about technology in general and AI in particular”
Experts still worry
Chinese AI researchers are serious about cyber, biosecurity, and loss-of-control risks. Some officials want tighter oversight, but not necessarily a joint slowdown with Washington.
“very concerned about cyber security and biosecurity risks”
The essence
A U.S.-China AI safety deal is more likely to start with narrow talks on weapons, misuse, and testing than with any broad slowdown.
