China and U.S. Map AI Cooperation Path Amid Strategic Rivalry

The geopolitical landscape around artificial intelligence has shifted from pure competition to a more complex reality where China and the United States maintain active dialogue channels while simultaneously racing for technological supremacy. This dual-track approach emerged from years of diplomatic friction, with both nations recognizing that certain AI risks transcend national boundaries.

According to China-US Focus, the two sides held their first inter-governmental AI dialogue in May 2024 in Geneva, Switzerland. The discussions centered on AI technology risks, global governance mechanisms, and issues of mutual concern. This meeting formally established artificial intelligence as a standing item in China-U.S. governmental dialogue, marking a cautious but meaningful step toward managing emerging technological risks through direct communication.

Subsequent developments reinforced this trajectory. In November 2024, Chinese President Xi Jinping and then-U.S. President Joe Biden met and reached an important consensus on ensuring nuclear weapons remain under human control. This shared understanding drew a clear red line regarding the potential militarization of AI, particularly in nuclear-related domains. The Los Angeles Times corroborates this accord, noting it was struck in Peru and represented a breakthrough moment where both sides acknowledged they could actually achieve something on AI governance.

The political transition following Donald Trump's return to office did not halt these efforts. After the 2025 leaders' meeting in Busan, both sides stated they would continue advancing mutually beneficial cooperation in AI. This signal suggests that regardless of which party controls the White House, Washington recognizes the necessity of maintaining engagement with Beijing on artificial intelligence-related issues.

Current discussions ahead of President Trump's state visit to China focus on reviving an emergency communication channel for AI matters. Officials told The Times that quiet discussions have taken place to explore this possibility, prompted by shared alarm over Anthropic's Mythos model. The model's capabilities are viewed across industry and government as those of an unprecedented cyberweapon, able to infiltrate and exploit digital communication systems including government databases, financial institutions, and healthcare programs.

Practical cooperation faces significant constraints. Nuclear weapons and other key strategic capabilities remain highly sensitive, making mutual transparency difficult to achieve in the near term. Challenges related to verification and persistent deficits in strategic trust in the arms control domain further complicate dialogue between the two sides. Chris McGuire of the Council on Foreign Relations notes that the Chinese government's willingness to make and abide by robust international commitments on AI safety is low, viewing such dialogues as opportunities to increase access to technology.

Despite these challenges, both nations have strong incentives to explore cooperation in less sensitive areas. On military-related issues, both sides could focus on promoting and deepening existing principled consensus rather than rushing into substantive military dialogues or technical cooperation. In non-military domains, considerable room exists for China and the U.S. to engage in dialogue and cooperation on assessing technological risks, addressing ethical concerns, and guiding AI development toward beneficial ends.

Track Two dialogue serves as an important supplementary channel. Think tanks, scholars, and retired military officers can continue discussions on risk assessment, ethical norms, and crisis decision-making. These efforts help sustain communication and build expert consensus that can eventually support official talks.

In non-military domains, China and the U.S. could work together to advance global governance of artificial intelligence. Both sides could focus on cross-border risks generated by AI and seek to promote risk-tiering, classification, and assessment frameworks acceptable to a broader range of countries. These governance challenges are shared by both countries, including risks related to loss of control over advanced AI systems, limited interpretability of large models, and AI-related biosecurity risks.

Approaching cooperation from a risk-based perspective may help reframe China and the U.S. from mutual "technological competitors" into "co-risk bearers." Even under conditions of strategic competition, both nations recognize that certain AI risks cannot be effectively addressed by any single country acting alone.

The consensus that nuclear weapons must remain under human control provides a solid foundation for the next stage of cooperation. President Trump is also trying to shape his image as a peace president, which opens space for both expansion and deepening of this consensus. China and the U.S. can encourage other nuclear powers, such as the U.K., France, and Russia, to adopt the same principle, gradually transforming a bilateral understanding into a multilateral one.

While a complete ban on AI in nuclear command, control, and communications systems is unrealistic, both sides should consider the potential benefits of AI-nuclear integration. Identifying mutually acceptable red lines remains crucial. The path forward requires balancing competition with cooperation, recognizing that the stakes of unmanaged AI risks extend far beyond bilateral relations to global security.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn