Chip design export controls: Chinese regulators are considering tightening export controls on semiconductor technologies, including potentially blocking TSMC and Qualcomm from manufacturing chips based on Chinese designs . This would help ensure that Chinese chip designs remain under the country's control.

Training data controls: Restrictions on training data being transferred outside of China are under discussion .

What would still be allowed: Overseas API access to Chinese AI services would likely remain permitted, even if model weight downloads are restricted .

Regulatory backdrop: In May 2026, the Cyberspace Administration of China published draft generative-AI export rules requiring security review for frontier-model exports to designated jurisdictions, with a 60-day consultation period .

The Moonshot AI Case and Escalating US Allegations

The US and China are in a rapid tit-for-tat escalation over AI controls, with the Trump administration simultaneously cracking down on both US companies (Anthropic) and Chinese firms (Moonshot AI).

On June 30, 2026, the US accused China's Moonshot AI of stealing technology from Anthropic's advanced AI model "Fable" to develop its Kimi K3 model . White House tech official Michael Kratsios publicly alleged a "large-scale" plot by Moonshot AI to appropriate Anthropic intellectual property .

On July 21, Treasury Secretary Scott Bessent warned that the Trump administration may impose sanctions on Chinese AI companies if it determines their open-source models were developed by appropriating US laboratory technology . "This administration endorses open-source models, but we do not condone intellectual property theft," Bessent said during a Fox Business interview .

The Trump Administration's Broader AI Crackdown

The US moves against Chinese AI are part of a wider pattern of aggressive export controls:

• In February 2026, Trump ordered federal agencies to phase out Anthropic technology, and the Pentagon classified Anthropic as a "supply-chain threat" . A federal judge temporarily blocked that order in March .
• In April 2026, the administration pledged a crackdown on Chinese firms "exploiting" US AI models .
• In June 2026, the US ordered Anthropic to deny foreign nationals access to its newest models, then partially lifted that export ban later the same month .
• As of July 22, 2026, the White House is reportedly considering banning Chinese AI models from the US entirely, with a Politico report quoting a White House official saying "any policy announcement would have to be paired with a robust enforcement mechanism" .

Timeline and Scope Uncertainty

Nothing is decided yet. Key unknowns include:

• China's timeline: The Ministry of Commerce consultations are ongoing with no announced deadline. The planned July 27, 2026 release of Moonshot's Kimi K3 open-weight model could be a flashpoint that accelerates or forces a decision .
• Regulatory form: It remains unclear whether China would adopt formal export control lists, licensing regimes, or softer "guidance"-based restrictions.
• Scope: Which models would be covered (frontier vs. all advanced), whether the rules apply retroactively to already-released open models, and how they interact with China's existing draft generative-AI export rules from May 2026 are all unresolved.
• US retaliation risk: The US is simultaneously weighing its own ban on Chinese AI models, and Treasury Secretary Bessent has made sanctions threats conditional on future US determinations of IP theft . The two countries' policy timelines are now tightly coupled, making the outlook highly unpredictable.

What This Means for the Global AI Landscape

If enacted, China's proposed controls would end the single most consequential asymmetry in the US–China tech contest: China's willingness to ship its best models to the world for free while Washington locks its own down .

For developers and companies that rely on Chinese open-weight models, the shift could mean restricted access to some of the most capable freely available AI systems. For US policymakers, it raises the question of how to respond to a world where both major AI powers are building walls around their most advanced technologies.

Chip design export controls: Chinese regulators are considering tightening export controls on semiconductor technologies, including potentially blocking TSMC and Qualcomm from manufacturing chips based on Chinese designs . This would help ensure that Chinese chip designs remain under the country's control.

Training data controls: Restrictions on training data being transferred outside of China are under discussion .

What would still be allowed: Overseas API access to Chinese AI services would likely remain permitted, even if model weight downloads are restricted .

Regulatory backdrop: In May 2026, the Cyberspace Administration of China published draft generative-AI export rules requiring security review for frontier-model exports to designated jurisdictions, with a 60-day consultation period .

The Moonshot AI Case and Escalating US Allegations

The US and China are in a rapid tit-for-tat escalation over AI controls, with the Trump administration simultaneously cracking down on both US companies (Anthropic) and Chinese firms (Moonshot AI).

On June 30, 2026, the US accused China's Moonshot AI of stealing technology from Anthropic's advanced AI model "Fable" to develop its Kimi K3 model . White House tech official Michael Kratsios publicly alleged a "large-scale" plot by Moonshot AI to appropriate Anthropic intellectual property .

On July 21, Treasury Secretary Scott Bessent warned that the Trump administration may impose sanctions on Chinese AI companies if it determines their open-source models were developed by appropriating US laboratory technology . "This administration endorses open-source models, but we do not condone intellectual property theft," Bessent said during a Fox Business interview .

The Trump Administration's Broader AI Crackdown

The US moves against Chinese AI are part of a wider pattern of aggressive export controls:

• In February 2026, Trump ordered federal agencies to phase out Anthropic technology, and the Pentagon classified Anthropic as a "supply-chain threat" . A federal judge temporarily blocked that order in March .
• In April 2026, the administration pledged a crackdown on Chinese firms "exploiting" US AI models .
• In June 2026, the US ordered Anthropic to deny foreign nationals access to its newest models, then partially lifted that export ban later the same month .
• As of July 22, 2026, the White House is reportedly considering banning Chinese AI models from the US entirely, with a Politico report quoting a White House official saying "any policy announcement would have to be paired with a robust enforcement mechanism" .

Timeline and Scope Uncertainty

Nothing is decided yet. Key unknowns include:

• China's timeline: The Ministry of Commerce consultations are ongoing with no announced deadline. The planned July 27, 2026 release of Moonshot's Kimi K3 open-weight model could be a flashpoint that accelerates or forces a decision .
• Regulatory form: It remains unclear whether China would adopt formal export control lists, licensing regimes, or softer "guidance"-based restrictions.
• Scope: Which models would be covered (frontier vs. all advanced), whether the rules apply retroactively to already-released open models, and how they interact with China's existing draft generative-AI export rules from May 2026 are all unresolved.
• US retaliation risk: The US is simultaneously weighing its own ban on Chinese AI models, and Treasury Secretary Bessent has made sanctions threats conditional on future US determinations of IP theft . The two countries' policy timelines are now tightly coupled, making the outlook highly unpredictable.

What This Means for the Global AI Landscape

If enacted, China's proposed controls would end the single most consequential asymmetry in the US–China tech contest: China's willingness to ship its best models to the world for free while Washington locks its own down .

For developers and companies that rely on Chinese open-weight models, the shift could mean restricted access to some of the most capable freely available AI systems. For US policymakers, it raises the question of how to respond to a world where both major AI powers are building walls around their most advanced technologies.