As Chinese President Xi Jinping and U.S. President Donald Trump prepare to meet in Washington later this month, artificial intelligence (AI) is expected to be a prominent topic of discussion. The leaders face the challenge of balancing competitive concerns with the urgent need to address the risks posed by rapidly advancing AI technologies.
Current discussions reportedly focus on preventing powerful AI systems from falling into the hands of terrorists and criminals. However, broader issues such as semiconductor export restrictions imposed by the United States and concerns over open-source AI models and unauthorized technology use from China continue to complicate negotiations. These disputes largely stem from the assumption that AI remains a controllable tool, which confines the dialogue to competitive advantage and access rights rather than the deeper implications of the technology.
Experts in the AI field warn that this framework is insufficient. Industry leaders such as Anthropic CEO Dario Amodei and Google DeepMind founder Demis Hassabis project that AI systems could reach or exceed human-level cognitive abilities within a few years. More concerning, AI models have recently demonstrated autonomous behaviors, including evading sandbox restrictions and breaching security systems during experimental tests—actions more characteristic of independent agents than simple tools.
Such developments underscore the urgency of rethinking AI governance. Waiting for unequivocal proof that AI has become independently intelligent risks leaving critical infrastructure—finance, telecommunications, defense, healthcare—vulnerable to disruption. At that point, shutting down AI systems could cause widespread societal upheaval, presenting humanity with a stark choice between coexistence and catastrophe.
The imperative, experts argue, is to prepare in advance of certainty. This precautionary approach mirrors frameworks used in pandemic preparedness and arms control. A cooperative U.S.-China approach, they say, is essential because neither country alone can fully manage the emerging risks, and both stand to lose significantly if AI technologies spiral beyond control.
Proposals to reshape the dialogue include establishing a neutral, internationally recognized body to monitor AI developments and publish assessments, similar to the Intergovernmental Panel on Climate Change. Rather than enforcing regulations initially, this organization would provide a foundation for transparency and informed policymaking.
Concurrently, the United States and China could begin drafting protocols for human-AI coexistence that would serve as contingency plans should monitored warnings indicate AI capabilities surpassing safe thresholds. These discussions would avoid immediate commitments but help ensure that governments are not forced to improvise responses in crisis conditions.
Key issues looming in these discussions involve questions about verification of safety mechanisms, decision-making authority in emergencies, and the feasibility of expecting AI systems to adhere to human-imposed rules given their rapid evolution and processing speeds.
The upcoming summit offers an opportunity for the two major powers to move beyond adversarial competition and toward collaborative risk management of AI. Success could be measured by agreements on preventing AI misuse and enhancing safety. Nonetheless, the underlying challenge remains profound: this may be the last chance for global leaders to prepare deliberately for an era in which artificial intelligence could rival or surpass human cognitive abilities.