Tech • AI • Robotics • Game
Kevin Rudd on the real risk of China and America’s AI race | The Economist
9.2/10TL;DR
The United States and China are opening an AI dialogue, but deep rivalry, mistrust and differing threat perceptions still stand in the way of urgent cooperation on advanced model safety.
KEY POINTS
A channel for talks is opening
A preliminary understanding to discuss artificial intelligence was reached in May, and that channel is now being activated. The development matters because the two countries dominate the most disruptive frontier of the technology and would be central to any meaningful guardrails on advanced systems.
Urgency remains contested
The core unresolved question is whether both governments share a common view of how urgent the risks from leading AI models have become. That includes risks from proprietary frontier models in the United States and from open-source development in China, especially as systems become more capable and harder to control.
Beijing’s public line stresses control
Xi Jinping has used language emphasizing human control over AI, a stance that can be read as both a policy signal and an attempt to project responsibility. By contrast, Donald Trump has been described as favoring a status quo approach while also framing AI as an arena of strategic competition.
Chinese confidence in cyber defenses may be overstated
A key view inside China appears to be that its cyber defenses are already strong, shaped by years of controlling unauthorized information flows inside a one-party state. That confidence rests partly on the belief that the country has mastered digital containment through extensive firewall and censorship capabilities. Critics of that view argue those tools may not withstand a broad AI-enabled attack on civilian, economic or even military infrastructure.
Technical communities are more alarmed than political systems
In both countries, a gap appears to exist between the technology sector and the policy establishment. Within China’s AI industry, concern has intensified as developments between the May Beijing meeting and the September Washington meeting accelerated at a pace and intensity that alarmed engineers and company leaders.
Autonomous agents are a central fear
The most specific concern is the emergence of advanced models that can generate autonomous agents able to act with limited human oversight. Those agents could spread through American or Chinese systems, adapt through deployment, and eventually be turned against the very states that helped create them.
China has not reached a settled internal consensus
Senior Chinese officials do not appear to have fully aligned around the same level of urgency seen among technical experts. The result is an internal tension between a traditional security mindset focused on regime stability and information control, and a newer recognition that frontier AI could create a wider systemic threat that existing political assumptions do not cover.
Strategic rivalry is the biggest obstacle
On both sides, the default position remains that AI is a live geopolitical contest in which victory delivers outsized power. The Chinese reading of repeated American calls for safeguards is colored by U.S. rhetoric about maintaining leadership and keeping China in second place, reinforcing fears in Beijing that any negotiation could lock in a disadvantage.
Arms-control logic complicates AI diplomacy
Chinese strategists have long argued in other domains, including nuclear policy, that the weaker side risks accepting a bad deal if it negotiates before reaching parity. That logic now appears to shape thinking on AI as well, making substantive limits difficult while China still sees itself as catching up in key areas.
The fear is that cooperation may require a crisis
The central dilemma is whether Washington and Beijing can act before a major AI-related disaster forces them to. Some analysts warn that without a severe real-world shock, the two powers may continue prioritizing competition over coordination even as the technology grows more capable and less predictable.
CONCLUSION
The opening of a bilateral AI channel is a notable step, but it remains far from a shared strategy. Unless both powers decide that frontier AI risks outweigh the race for dominance, meaningful safeguards may come only after a serious crisis.
Ask a question