The conventional wisdom in Washington assumes the United States and China are locked in an existential race to reach artificial general intelligence (AGI), with the winner gaining permanent strategic dominance. But according to a new analysis from the Quincy Institute for Responsible Statecraft, this framing rests on three deeply flawed assumptions that are now driving nearly a trillion dollars in annual capital spending and reshaping U.S. foreign policy.

Is There Actually a "Finish Line" in AI Development?

The core problem with the AI race narrative is the belief that crossing some decisive threshold in AI capability will grant one nation permanent advantage over the other. But AI development doesn't work that way. Unlike nuclear weapons, which are discrete, physically securable, and difficult to replicate, AI models are software that can be copied and improved upon rapidly across borders.

The evidence suggests capability gaps between the U.S. and China are narrowing faster than many policymakers realize. What was once estimated as a 12 to 14-month performance gap between American and Chinese frontier AI models has shrunk to roughly 2 to 3 months, even as Washington has steadily ramped up export controls on advanced semiconductors. On OpenRouter, the largest model-routing platform, Chinese open-weight models have surged from under 2 percent of token traffic in late 2024 to holding the top six positions among the most-used models as of July 2026.

Recent releases from Chinese companies like DeepSeek and Moonshot AI are now competitive with leading U.S. models from OpenAI and Anthropic, often at a fraction of the cost. DeepSeek's V4 Flash trails OpenAI's GPT-5.6 Luna by just a single point on the Artificial Analysis Intelligence Index, while costing roughly 60 percent less per task even after OpenAI's recent 80 percent price cut.

What Are the Real Security Threats From Advanced AI?

The analysis argues that focusing exclusively on nation-state competition misses the actual security risk. The most dangerous AI models may not be the largest frontier models built by governments or major tech companies, but rather smaller, specialized models that could be created or weaponized by non-state actors for deploying biological, chemical, or cyber weapons.

This distinction matters because it suggests the path forward should involve global cooperation on safety standards rather than maximum denial and competition. The upcoming summit between President Trump and President Xi Jinping presents an opportunity to establish sensible cooperative measures around shared harm and safety standards, according to the analysis.

How to Build International AI Cooperation Instead of Competition

Rather than pursuing an unwinnable race, the analysis proposes a three-step framework for establishing global AI safety measures:

• Unilateral Commitment: Individual companies like Anthropic could give third-party evaluators employee-like access to verify safety practices, report incidents, and assess alignment of models and training pipelines, with the right to publish findings without company editorial control.
• Democratic Coordination: Frontier AI companies in democratic countries would coordinate common safety standards and limits on unchecked progress through regulation and voluntary standards.
• Global Coordination: The U.S. and its allies would attempt cooperation to establish broader international constraints on frontier AI, while addressing the major challenge of verifying compliance.

The analysis also proposes different layers of possible agreement with China, ranging from easiest to hardest. These include banning AI for bioweapons development, mutual pre-release testing via a global standards body, speed limits on recursive self-improvement analogous to nuclear arms control treaties, and a full pause on development.

However, the gap between this vision and Beijing's position remains significant. Chinese Foreign Ministry spokesperson Guo Jiakun responded to calls for slowing down by noting that fearmongering and confrontation would hinder global AI cooperation, adding that China emphasizes both development and security equally.

Why Is the Industry Divided on Pacing AI Development?

The debate over how fast to develop frontier AI has split both the technology industry and the U.S. government. Anthropic CEO Dario Amodei called for slowing development to ensure safety can keep pace with capability advances, and his proposal has gained support from OpenAI CEO Sam Altman, SpaceX AI CEO Elon Musk, and Google DeepMind Chair Demis Hassabis.

Leading AI companies including Anthropic, Google, and OpenAI have reportedly discussed establishing an independent industry standards body to test powerful AI models before public deployment, modeled after the Financial Industry Regulatory Authority. However, not all industry leaders agree. Meta CEO Mark Zuckerberg wrote that companies can slow down without an industry-wide pact, noting Meta had delayed its AI agent Muse to strengthen safety.

The Trump administration has rejected calls for slowing down, with President Trump calling concerns about AI safety a "hoax" and arguing that the U.S. must continue developing AI rapidly because "whoever wins AI, wins". Vice President JD Vance noted that responsibility for regulating AI rests with developers rather than government.

Meanwhile, Congress has introduced bipartisan bills aimed at increasing oversight, including the Stop Rogue AI Act, which would give federal agencies tools to identify and shut down dangerous AI systems, and the AI Kill Switch Act, which would require developers to build capabilities to slow, suspend, or shut down systems. However, reports indicate nearly all major regulatory AI proposals remain stalled in Congress.

The European Union is positioning itself as a third pole in this debate. EU officials met in Brussels on September 17 to discuss rules for new AI models, and European Commission President Ursula von der Leyen called for cooperation with Canada and the UK on model evaluation, verification, early warning, and AI security.

As preliminary U.S.-China AI safety negotiations proceed ahead of Xi Jinping's planned visit to Washington, the fundamental question remains unresolved: whether the greater risk from advanced AI lies in moving too quickly or too slowly, and who should decide where that line lies.