Washington is obsessing over the wrong crisis.
When diplomatic summits between the United States and China are framed around an impending artificial intelligence catastrophe, policy analysts, defense contractors, and tech executives treat it like a modern Cuban Missile Crisis. They warn that without immediate bilateral guardrails, runaway algorithms and autonomous weapons will trigger an accidental war.
That narrative is convenient for bureaucrats. It is also fundamentally flawed.
Framing a high-stakes diplomatic meeting around an "AI crisis" relies on a lazy consensus. It assumes that artificial intelligence is an independent, existential threat that Washington and Beijing can mutually regulate through handshakes and non-proliferation treaties. In reality, treating AI as a standalone diplomatic emergency obscures the actual friction points driving U.S.-China relations—and actively undermines Western technological competitiveness.
The Myth of the Bilateral AI Treaty
The dominant foreign policy establishment loves a Cold War analogy. They look at nuclear arms control treaties from the 1970s and assume the same playbook applies to frontier models and compute clusters.
It does not. Nuclear weapons required massive, easily detectable physical infrastructure—uranium enrichment facilities, missile silos, and heavy water reactors. Satellites could verify compliance from orbit.
Artificial intelligence is software running on commercial data centers. You cannot verify a bilateral pause on model training without intrusive, real-time inspection of private server farms and proprietary codebases. Neither Washington nor Beijing will ever grant foreign inspectors root access to their sovereign compute infrastructure. Any treaty signed during a state visit claiming to "halt dangerous AI development" is pure political theater. It creates an illusion of safety while both nations continue building behind closed doors.
I have watched enterprise tech leaders and defense policymakers burn tens of millions of dollars preparing for international compliance frameworks that do not exist and cannot be enforced. While Western executives stall deployment waiting for regulatory clarity from diplomatic summits, their competitors are shipping code.
Compute Is the Real Chokepoint, Not Code
To understand why the "AI crisis" framing fails, you have to look at the physical layer of the technology stack.
Foreign policy pundits love talking about alignment, model weights, and algorithmic safety. Engineers know that models are merely mathematical abstractions until they meet silicon. The true geopolitical battleground is not a diplomatic summit in Washington; it is the semiconductor supply chain stretching from Dutch lithography machines to Taiwanese fabrication plants.
| Diplomatic Narrative | Reality on the Ground |
|---|---|
| Focus: Joint AI safety protocols and ethical pledges. | Focus: Export controls on extreme ultraviolet (EUV) lithography and high-bandwidth memory. |
| Assumption: Both nations fear an uncontrollable superintelligence. | Assumption: Both nations want asymmetric economic and military advantage through compute dominance. |
| Outcome: Non-binding communiqués and working groups. | Outcome: Accelerated domestic chip manufacturing and supply chain decoupling. |
When analysts argue that an AI crisis heightens the stakes of a bilateral visit, they get the causality backward. The visit does not solve the AI race; the structural reality of the semiconductor market dictates the limits of the visit. China does not need Washington's permission to write open-weights models, and Washington will not trade away its export controls on advanced GPUs in exchange for vague promises about algorithmic safety.
Why "AI Safety" Serves Strategic Interests
If bilateral AI treaties are unenforceable, why do both sides keep putting AI on the diplomatic agenda? Because strategic ambiguity benefits both capitals, though for entirely different reasons.
For Beijing, participating in global AI governance forums projects the image of a responsible stakeholder while buying time for domestic semiconductor firms like SMIC and Huawei to close the hardware gap. By engaging in endless dialogues about "human-centric AI" and risk mitigation, Chinese diplomats slow down unified Western action on trade and technology restrictions.
For Washington, hyping an existential AI crisis justifies sweeping industrial policy and consolidates regulatory power among a handful of entrenched American incumbents. When major AI labs testify before Congress that their technology could end civilization unless strictly controlled, they are not begging to be stopped. They are asking the government to build a regulatory moat that smaller, open-source competitors cannot cross.
The Trap of Regulatory Capture
Look closely at who benefits when summits focus on "existential AI risks." It is rarely the open-source developer community or mid-sized enterprise innovators.
- Incumbent Tech Giants: Gain massive influence over the standards-setting bodies that emerge from diplomatic talks.
- Defense Consultants: Secure lucrative contracts to model hypothetical AI conflict scenarios.
- Bureaucracies: Expand their mandates into international tech governance without having to solve harder, immediate problems like intellectual property theft or market access.
There is a legitimate downside to dismissing safety dialogues entirely. Maintaining open military-to-military communication channels—specifically keeping humans in the loop for nuclear command and control—is basic common sense. Nobody wants an automated sensor glitch mistaken for a first strike.
However, conflating basic nuclear command hygiene with broad "AI crisis diplomacy" is a category error. You do not need a grand bargain on artificial intelligence to agree that automated systems should not launch intercontinental ballistic missiles.
The Questions Analysts Should Actually Be Asking
If you want to understand what actually matters when Chinese leadership visits Washington, stop reading press releases about AI safety working groups. Look at the structural economics that neither side can negotiate away.
Can export controls actually stop Chinese AI progress?
No, they can only slow it down and increase its cost. Export restrictions on top-tier NVIDIA hardware force Chinese engineers to optimize software efficiency, rely on distributed clusters of older chips, and pour state subsidies into domestic alternatives. Tightening the screws creates short-term friction, but over a five-to-ten-year horizon, it guarantees a fully independent, non-Western tech stack.
Is open-source AI a national security threat?
The establishment consensus says yes, arguing that releasing model weights hands adversaries dangerous capabilities on a silver platter. The reality is the exact opposite. When American companies and researchers lead in open-weights models, the global developer ecosystem builds on Western standards, Western tooling, and Western cloud infrastructure. If Washington restricts open-source development out of fear, developers in Southeast Asia, Europe, and Latin America will simply adopt open models released by Chinese tech giants like Alibaba and DeepSeek. Surrendering the open-source ecosystem is unilateral technological disarmament.
What happens if both sides sign an AI declaration?
Nothing changes in the labs. Compute clusters keep spinning, talent keeps getting poached, and military integration of autonomous drones and electronic warfare continues unabated. A diplomatic communiqué is a press event, not an engineering constraint.
Stop Negotiating Guardrails and Start Building Leverage
Policymakers need to abandon the fantasy that diplomacy can tame technological disruption. History shows that transformative technologies—from electricity to the internet—are governed by whoever builds the infrastructure and sets the commercial standards, not by diplomats drafting preventative treaties.
If Washington wants real leverage in its relationship with Beijing, it must shift from defensive containment to aggressive execution:
- Fix the domestic energy grid. Frontier AI is bottlenecked by gigawatts, not diplomatic agreements. Permitting reform for nuclear, natural gas, and grid transmission matters infinitely more than a joint statement on AI ethics.
- Protect the open-source ecosystem. Keep frontier research open so global software standards remain anchored in democratic markets rather than state-controlled ecosystems.
- Focus diplomacy on hard red lines. Keep bilateral talks strictly limited to verifiable military deconfliction—specifically prohibiting autonomous integration into nuclear command systems—and drop the theatrical discussions over commercial model alignment.
Diplomatic summits are designed to manage friction, not rewrite the laws of technological diffusion. As long as Washington treats artificial intelligence as a diplomatic crisis to be negotiated away rather than an industrial race to be won, it will keep losing ground while celebrating empty communiqués.
