Distrust Blocks US-China AI Safety Cooperation

The United States and China are both grappling with the safety implications of their rapid progress in artificial intelligence, yet the very factors driving this progress are blocking the cooperation experts say is necessary to manage the risks. As the technology becomes a central pillar of the rivalry between Washington and Beijing, the mutual distrust surrounding it threatens to derail efforts to head off the greatest dangers posed by these systems.

This tension is set to dominate discussions at a summit in Washington next week between President Donald Trump and Chinese leader Xi Jinping. While some U.S. researchers urge companies to slow the pace of development to address safety concerns, political leaders remain divided on how to balance national security with global stability.

Competing Narratives on Safety

In the U.S., prominent AI leaders and researchers argue that recent incidents involving autonomous agents demonstrate the need for caution. They suggest that the industry is moving too fast, potentially outpacing the ability to control the technology. However, this view is not universally shared within the American political establishment.

President Trump has dismissed these safety warnings as a hoax, prioritizing instead the maintenance of American dominance over China through export controls. This creates a disconnect between the technical community’s safety concerns and the strategic imperative to outpace a geopolitical rival. The result is a policy environment where safety measures are often viewed through the lens of competitive advantage rather than shared human benefit.

China Sees Slowing as Sabotage

Beijing interprets calls to slow AI development as targeted efforts to hinder its own progress. Chinese officials argue that the U.S. is using safety rhetoric to mask a desire to stave off competition. They point to accusations of illicit distillation, a process where companies train their models using outputs from rival systems, as evidence of a zero-sum mindset in Washington.

While U.S. companies claim Chinese firms are improperly extracting knowledge from their advanced models, Chinese counterparts maintain that distillation is a standard industry practice. This mutual accusation has hardened positions, making it difficult for either side to view the other’s intentions with goodwill. As reported by GN technics/ai (en-US), this cycle of blame reinforces the perception that cooperation is impossible without first conceding competitive ground.

Zero-Sum Thinking Blocks Dialogue

Experts warn that the U.S. approach is fundamentally flawed if it assumes that one side winning the AI race means the other must lose. Francesca Giovannini, a director at the Harvard Belfer Center, notes that this zero-sum theory prevents the U.S. from conceding anything that might benefit China. Without a willingness to share benefits, meaningful dialogue on risks remains stalled.

China, which views its open-weight AI models as a tool for diplomatic influence, is unlikely to agree to any slowdown dictated by U.S. corporate interests. Analysts suggest that Beijing sees this moment as a historic opportunity to assert its technological independence. Until both sides move beyond the notion that AI is a weapon to be won or lost, the risk of uncoordinated development will remain a significant threat to global stability.