How China's Proposed AI Export Controls Mirror the U.S. Playbook

On July 7, 2026, Reuters reported exclusively that Chinese authorities have held meetings with top tech firms over the past month about potentially restricting overseas access to China's most advanced AI models, including systems not yet released ^SAR. The meetings, led by China's Ministry of Commerce, involved Alibaba, ByteDance, and Z.ai ^ARL. Officials indicated that leaks or theft of proprietary AI technology could become punishable under national security law ^SRF.

The Proposed Tiered Control Framework

The reported approach would introduce tiered restrictions on access to frontier AI models:

• Most advanced / frontier models — those closest to the technological frontier — would face the tightest restrictions, potentially barred from foreign access outright ^ARL.
• Less capable or older models could remain available overseas, mirroring a graduated control logic ^AR.
• Some open-weight releases also could be affected by the new restrictions ^AER.

The framework builds on China's May 28, 2026 draft generative-AI export rules, published by the Cyberspace Administration of China (CAC). Those draft rules require a security review for frontier-model export to designated jurisdictions, and establish their own three-tier classification for generative-AI exports: unrestricted for small-scale models below threshold parameter counts; restricted for frontier models; and a middle tier requiring case-by-case review ^AT.

These model-level controls are part of a broader clampdown. In May 2026, Bloomberg reported that China restricted overseas travel for top AI talent at private firms like Alibaba and DeepSeek, requiring government approval before trips abroad — a policy shift from prior notification to mandatory prior approval ^TAB. And on July 1, 2026, China imposed broad national security rules on overseas investments, extending curbs beyond goods and data to cover the export of services through sending technical experts abroad or carrying out training overseas ^GSF.

How These Measures Mirror U.S. Export Controls

The symmetry between the two countries' approaches is striking:

U.S. Approach (Biden-era)	China's Mirror Measures
AI Diffusion Rule (Jan 2025): Three-tier global classification for AI chip and model-weight exports TNC. Tier 1 (allies) free access; Tier 2 (most countries) annual caps; Tier 3 (China, Russia) barred outright TEF.	Proposed tiered model access: Frontier models restricted or barred; graduated access for less-capable models ARL.
Model-weight controls (Jan 2025): First-ever export controls on AI model weights — treating frontier weights as sensitive as semiconductor wafers SSA.	Draft generative-AI export rules (May 2026): Security review required before exporting frontier models AT.
Entity list designations: Adding Chinese AI/tech firms to restricted entities for national security reasons SB.	National security law for AI leaks: Officials suggested leaks/theft of AI could be treated as national security crimes SCR.
Chip controls targeting compute access: Restricting advanced chip sales to China SFA.	Domestic chip mandate: China banned foreign AI chips from state-funded data centers RS.
BIS license requirements for chip/tech exports SB.	MOFCOM-led licensing / security reviews for AI model exports ATR.

The critical difference: The U.S. Biden-era AI Diffusion Rule was rescinded by the Trump administration in May 2025 before it took full effect ^EFA, while China is actively building out its control architecture based on the same playbook.

Broader Trend: The Bifurcation of Frontier AI

These developments signal a clear and accelerating trend in superpower competition:

1. Mutual technology denial is becoming the new normal. Both the U.S. and China are now treating frontier AI model weights and frontier AI talent as strategic assets that must be denied to the other side ^ACSR.

2. The AI stack is being severed at every layer. Controls now span: chips (U.S. blocks advanced chips to China; China mandates domestic chips for state data centers) ^SSR, model weights (U.S. Jan 2025 rule; China's May 2026 draft rules) ^ST, talent (China restricts AI researcher travel) ^TAB, and investment (China's new national security investment rules) ^GSF.

3. A "bifurcated global AI ecosystem" is emerging. Rather than a single global market for frontier AI, the world is splitting into U.S.-aligned and China-aligned technology blocs, each with its own chip supply chains, model ecosystems, and regulatory standards ^ESM.

4. Beijing is filling the vacuum left by Washington's policy reversals. The U.S. pioneered AI export controls (2022-2025), but with the Trump administration rolling back some restrictions ^EFB, China is developing its own export control apparatus modeled on the U.S. playbook ^TA. As Reuters noted in February 2026: "As Trump reins in China tech curbs, Beijing's export controls come of age" ^R.

5. The competition is existential. Both sides frame frontier AI as a matter of national security, military advantage, and economic supremacy ^STSR. The MERICS report (June 2026) notes that "Beijing has made 'independent and controllable' AI a key objective" in response to U.S. export controls ^M.

In short, the U.S.-China AI rivalry has entered a reciprocal technology denial phase — each side building walls around its most advanced AI capabilities, mirroring the other's tactics, and accelerating the fragmentation of global AI governance into competing spheres of influence.