US AI Illustration: Chen Xia/GT

Should the development of artificial intelligence (AI) in the US accelerate or slow down? A heated debate has erupted over this issue between the US government and industry figures, led by Anthropic CEO Dario Amodei. 

However, what deserves more scrutiny than the pace is the direction. 

When accelerating is seen as outpacing China, and decelerating is designed to maintain an advantage over China, both positions are driven by the same Cold War mentality. Framing AI development as a zero-sum game and viewing China's progress as a threat to US security - no matter how skillfully the pace is adjusted - will hardly lead to technology for the greater good or lasting security.

The US' executive order regarding AI has listed enhancing the global AI dominance of the US as a policy goal, linking it to economic competitiveness and national security. Within this framework, accelerating innovation is easily imbued with strategic implications that transcend industrial development: Technological leadership must be translated into rules-setting dominance, and competitive advantages must be solidified into international power.

Why is the "Chinese overtaking the US" directly equated with a security threat? Cold War mentality presumes that China's progress inevitably harms US interests, and that a narrowing of the technological gap between China and the US necessarily implies increased danger. Competitors are thus cast as potential adversaries, normal technological development is imbued with geopolitical hostility, and the security narrative slides toward the demonization of other countries.

Truly effective security assessments should focus on verifiable capabilities, uses and behaviors, applying consistent standards. When the same technology is labeled an "opportunity" in one country but a "threat" in China, security discussions devolve into double standards.

Acceleration aimed at overwhelming an adversary will intensify the pressure to deploy first, making it easier for competitive anxiety to supplant prudent assessment; deceleration aimed at freezing an advantage, on the other hand, will struggle to secure equal participation and sustained trust. Both paths are prone to falling into the same cycle: The more one fears being surpassed, the more one emphasizes exclusive control; the more control is strengthened, the deeper the mistrust becomes, and the harder it is to achieve true security.

Therefore, correcting the course is a priority. The US should not make the suppression of China's development a prerequisite for the success of its AI governance. 

The existence of competition between China and the US does not mean that the two must head toward full-scale confrontation. 

The leaders of China and the US have already provided strategic guidance for overcoming this inertia of confrontation. On May 14, the two leaders agreed to establish a constructive relationship of strategic stability between China and the US as the new framework for bilateral relations, featuring positive stability with cooperation as the mainstay, healthy stability with competition within proper limits, constant stability with manageable differences and lasting stability with expectable peace. In implementing this consensus, AI should serve as a key area for testing whether the two sides are moving toward each other. 

The key lies in translating political will at the head-of-state level into sustainable communication and action. Both sides can engage in professional exchanges on the security assessment of cutting-edge models, the sharing of information on major risks and the prevention of malicious misuse, while enhancing necessary transparency and protecting trade secrets and legitimate security interests. Moreover, they can support research institutions in collaborating on security technologies and applications that serve the public interest.

At the same time, China-US cooperation should be open to broader international participation, support governance discussions within the UN framework, and prioritize capacity building and the legitimate concerns of developing countries. 

What the US really needs to overcome in its "speed debate" over AI is the cognitive barrier erected by Cold War mentality. If the starting point is skewed and the direction is wrong, neither acceleration nor deceleration can bring lasting security. To move beyond the pan-securitization anxiety over "China overtaking" and advance the sound development of AI within the framework of constructive strategic stability is the responsible choice for the peoples of both countries and the future of the world.

The author is director of the International Research Institute of Global Cyberspace Governance at Fudan University. opinion@globaltimes.com.cn