With artificial intelligence (AI) increasingly shaping global technology and security landscapes, the upcoming summit in Washington between U.S. President Donald Trump and China’s President Xi Jinping is set to include discussions on AI safety amid rising geopolitical tensions. The United States and China, the world’s foremost AI competitors, approach AI safety from fundamentally different perspectives, complicating efforts toward a cooperative dialogue.

In the U.S., concerns focus largely on preventing the emergence of autonomous AI systems that could operate beyond human control or be used by rogue actors to develop new weapons. While President Trump has minimized the necessity of stringent regulatory measures, leading figures in the American AI sector advocate for a cautious slowdown to prioritize safety considerations. U.S. AI developers such as OpenAI and Anthropic typically restrict access to their most advanced models via paid gateways, reflecting an attempt to manage risks associated with open availability.

Conversely, China’s AI safety concerns predominantly center on protecting the political dominance of the Communist Party. Chen Yixin, China’s minister of state security, recently highlighted risks such as the technology’s potential to facilitate the rapid spread of disinformation, conduct “cognitive warfare,” and incite social unrest—all threats to the state’s political-security environment rather than humanity at large. Chinese AI firms like DeepSeek often release open-weight AI models publicly, a strategy intended to promote Chinese AI’s global influence and counterbalance U.S. containment efforts.

This divergence in priorities poses challenges for meaningful collaboration. Former U.S. National Security Council official Chris McGuire, now a senior fellow at the Council on Foreign Relations, noted that Beijing has consistently declined serious cooperation proposals with the U.S. on AI safety. McGuire emphasized that China’s primary goal is achieving global AI leadership and that it is unlikely to impose development constraints as long as it remains slightly behind U.S. labs.

Despite the sharp differences, some diplomatic groundwork has been laid. Trump and Xi agreed on initiating AI dialogues during their May summit in Beijing, though substantial progress has been limited. Ahead of Xi’s Washington visit, U.S. Treasury Secretary Scott Bessent is scheduled to meet China’s top economic official, He Lifeng, in New York to prepare for AI discussions. Bessent affirmed the U.S. position as AI leader and expressed openness to talks on preventing shared risks and system bifurcation.

On the Chinese side, leading party figures have conveyed mixed signals. Wang Huning, a senior Communist Party ideologue, reportedly acknowledged the importance of bilateral AI guardrail conversations, yet analysts suggest China may underestimate certain threats. Venture investor Gary Rieschel warned that China’s reliance on the “Great Firewall,” a sophisticated internet control system, does not adequately defend against AI-driven cyber intrusions, which can automatically seek vulnerabilities—a type of threat beyond traditional content censorship.

Chinese officials have shown strong interest in the status of AI talks and the potential easing of U.S. technology export restrictions. U.S. officials visiting China describe a more confident Chinese stance regarding their AI capabilities, asserting technological parity and promoting open access to domestic AI models. However, they have been denied access to key Chinese AI research facilities, reflecting persistent opacity.

Previous attempts to foster formal bilateral AI dialogues, including under the Biden administration, have faltered due in part to Beijing’s decision to vest talks within the foreign ministry rather than technical agencies, limiting substantive exchange. U.S. officials observe that the seriousness of forthcoming AI negotiations may hinge on whom China appoints to lead the discussions.

Ultimately, the fundamental difference remains: the U.S. aims to develop AI safeguards designed to protect society broadly, whereas China prioritizes safeguards that preserve the Communist Party’s control. This fundamental mismatch complicates prospects for a durable and comprehensive agreement on AI safety between the two powers.