Trump Rules Out US-China AI Joint Venture Over Technology-Sharing Concerns

President Donald Trump said on September 29, 2026, that he did not want the United States to pursue a joint artificial intelligence venture with China because such an arrangement could expose sensitive U.S. technological knowledge.

Trump made the comments while launching America.gov at the Andrew W. Mellon Auditorium in Washington, D.C. The White House event record confirms the setting, while a transcript of Trump's remarks from Roll Call Factbase records him referring specifically to a "joint venture" and saying it could involve "giving a lot of secrets" to China.

The distinction between a joint venture and broader AI cooperation is critical. Trump did not announce that the United States was ending every form of AI dialogue, safety cooperation, research contact or diplomatic engagement with China.

Four days earlier, the White House had announced a bilateral U.S.-China dialogue on AI-related risks and benefits, along with a communications channel for AI incidents. China's Foreign Ministry subsequently described the same arrangements in its official list of bilateral understandings.

The available evidence therefore points to a narrower policy distinction: Trump rejected the idea of jointly developing AI in an arrangement he believed could expose U.S. technology, while the two governments continue to maintain channels for discussing AI risks and incidents.

Key Takeaways

• Trump said on September 29, 2026, that he opposed a U.S.-China AI joint venture because it could give China access to U.S. technological knowledge.

• His remarks concerned a joint venture or shared-development concept, not a publicly announced prohibition on all AI cooperation.

• The United States and China had just established a bilateral AI risk-and-benefit dialogue and an AI incident-communication channel.

• Risk communication and AI safety dialogue do not inherently require sharing proprietary models, source code, model weights or advanced semiconductor technology.

• Current BIS rules still impose targeted licensing requirements on specified advanced-computing transactions involving China-linked entities.

• Treasury's outbound-investment rules continue to apply to specified AI, semiconductor and quantum transactions involving covered China-related entities, while new COINS Act regulations are developed.

• Businesses should evaluate cross-border AI activity according to the specific technology, transaction, parties, ownership, destination and end use rather than treating all AI activity as legally equivalent.

What Is Confirmed, and What Is Not?

What Did Trump Actually Say About AI Cooperation With China?

Trump's remarks came during his September 29 launch of America.gov, an AI-powered federal services platform. The White House has published an official video of Trump's September 29 remarks and a separate fact sheet on the America.gov launch.

While discussing technological competition with China, Trump said somebody had suggested: "let's do a joint venture. Let's work together with China." He rejected that approach because, he argued, it could provide China with access to U.S. technological knowledge.

Trump did not identify the person or group that proposed the idea. He also did not name potential companies, define an ownership structure, describe technologies that would have been involved, or say that a formal government-to-government joint venture had been negotiated.

That is why "joint venture" is more accurate here than the much broader term "AI cooperation."

Trump also acknowledged that U.S.-China engagement continues in other areas. According to the transcript, after rejecting the joint-development concept he said the United States works with China "in so many ways."

A Short Timeline of the U.S.-China AI Discussions

Ahead of the Xi meeting, Reuters reported that Trump said AI would be a major topic of discussion.

China's account of the September 24 talks said Xi argued that the two countries could continue AI dialogue, discuss risks and benefits and work against misuse. The same Chinese Foreign Ministry account attributes to Trump the view that the countries should maintain dialogue and strengthen AI cooperation.

The U.S. and Chinese descriptions use slightly different terminology. The White House now calls the mechanism the U.S.-China Super Intelligence Dialogue, while China's Foreign Ministry describes it as the China-U.S. AI Dialogue. Both governments say its purpose includes exchanging views on risks and benefits, and both say another exchange is expected in November 2026.

Why Is Sharing AI Technology With China So Sensitive?

"AI technology" is not one uniform category.

A consumer chatbot, an ordinary software application, a frontier model's trained weights, a cluster of advanced GPUs and semiconductor manufacturing technology can raise very different commercial, regulatory and national-security issues.

For frontier AI development, strategically important assets can include advanced computing chips, model weights, specialized training infrastructure, proprietary algorithms, non-public optimization techniques, cybersecurity capabilities and technical know-how.

Model weights deserve particular attention. They contain the learned parameters produced during model training. Access to the weights of a highly capable proprietary model can therefore provide access to capabilities that required substantial computing resources, data and expertise to develop.

Current U.S. export controls reflect this differentiated approach. The Bureau of Industry and Security's current Export Administration Regulations contain controls affecting advanced computing and specified advanced AI-related technology.

More specifically, BIS issued guidance on May 31, 2026 concerning advanced-computing exports. It states that licensing requirements continue to apply to specified advanced-computing items for entities headquartered in Country Group D:5 or Macau, or whose ultimate parent is headquartered there, even when the immediate entity is located elsewhere. Country Group D:5 includes China.

That does not mean every AI system or transaction involving China is prohibited.

Export-control obligations depend on the specific item, classification, parties, destination, ownership, end use and available authorization or license exception.

The Bigger U.S.-China AI Competition

Trump's remarks sit within a broader technological competition involving frontier models, advanced semiconductors, data-center infrastructure, research, robotics and commercial AI deployment.

The competition is also difficult to reduce to a simple claim that one country has decisively "won."

Stanford's 2026 AI Index Report found that U.S. and Chinese models had traded positions at the top of performance rankings since early 2025. As of March 2026, the report measured only a 2.7% gap between the leading U.S. and Chinese models on the benchmark it examined. Stanford also found different strengths across the two countries, with the United States producing more top-tier models while China led in measures including publication volume and patent output.

Advanced computing is particularly important because training highly capable models can require large amounts of specialized semiconductor capacity.

That is one reason U.S.-China AI competition extends beyond software. AI chips, semiconductor manufacturing, cloud infrastructure, data centers and technical knowledge can all become relevant to trade and national-security policy.

AI Risk Cooperation Is Different From Technology Sharing

This is the most important distinction in interpreting Trump's comments.

Governments can cooperate on AI risk without jointly developing proprietary technology.

They can maintain incident hotlines, discuss misuse scenarios, exchange views on safety, develop international norms or communicate during an AI-related crisis without transferring model weights, source code, semiconductor manufacturing techniques or confidential research.

That distinction is visible in the September U.S.-China agreements.

The White House's September 25 fact sheet says the countries created a dialogue to exchange views on AI-related risks and benefits and agreed to establish a bilateral communication channel for incidents.

Nothing in that announced arrangement says the United States will give China proprietary American models, source code, model weights or advanced semiconductor technology.

A joint venture would be materially different. Depending on its structure, it could involve shared capital, ownership, personnel, infrastructure, intellectual property, development work or operational control.

Trump's comments focused on that deeper form of technological integration.

How Trump's Comments Fit Into Current U.S.-China AI Policy

Current U.S. policy combines selective technological restrictions with continued international engagement.

One source of confusion is the Biden-era AI Diffusion Rule. In May 2025, the Commerce Department announced that it was rescinding that framework before its new compliance requirements were enforced. The BIS announcement on the AI Diffusion Rule also said Commerce was taking separate steps involving semiconductor export controls.

The rescission did not eliminate all China-related advanced-computing restrictions.

BIS clarified that point directly in its May 2026 guidance, explaining that a pre-existing licensing requirement continues to apply to relevant advanced-computing items for entities headquartered in Country Group D:5 or Macau, including where an entity's ultimate parent is headquartered there.

The current EAR also contains specific advanced-computing rules, including provisions addressing certain transactions based on an entity's headquarters or ultimate parent. EAR Part 744 provides the current regulatory text for these end-user and end-use controls.

The result is a targeted framework rather than a universal prohibition on AI trade with China.

Outbound Investment Adds Another Compliance Layer

Technology exports are only part of the picture.

The U.S. Treasury Department also administers an outbound-investment regime addressing specified U.S. investments connected to semiconductors and microelectronics, quantum information technologies and artificial intelligence.

According to the Treasury Department's Outbound Investment Security Program, China, Hong Kong and Macau are designated countries or jurisdictions of concern for the program. The original rules took effect on January 2, 2025.

Congress subsequently enacted the Comprehensive Outbound Investment National Security Act of 2025, known as the COINS Act, as part of the fiscal 2026 National Defense Authorization Act.

Crucially, Treasury says in its current outbound-investment FAQs that the existing rules at 31 CFR Part 850 remain in force until Treasury issues regulations implementing the COINS Act.

Treasury also makes clear that the regime does not prohibit all investment involving China. It targets specified covered transactions involving particular technologies and entities.

What This Could Mean for the Global AI Industry

The immediate significance of Trump's remarks is more political and strategic than regulatory. His statement itself did not create a new export-control rule or announce a new legal prohibition.

However, it indicates that proposals involving shared U.S.-China development of sensitive AI capabilities may face substantial political and national-security scrutiny.

For technology companies, greater separation in selected parts of the U.S. and Chinese AI ecosystems could affect access to advanced computing, cloud infrastructure, research relationships and cross-border investment.

For semiconductor and data-center businesses, compliance can increasingly depend not only on where an item is physically shipped but also on the headquarters and ultimate ownership of the entity receiving it. BIS's May 2026 advanced-computing guidance illustrates that point directly.

Universities and research organizations can face a different set of questions. Open academic publications, fundamental research, proprietary technical data, controlled hardware and private model weights should not automatically be treated as the same thing.

International AI governance may therefore develop along two tracks: restrictions around strategically sensitive technology and continued engagement around shared risks.

Those approaches are not inherently contradictory.

What This Means for Businesses Using AI

Businesses should not translate Trump's comments into a blanket assumption that U.S.-China AI activity is now prohibited.

A more useful approach is transaction-specific.

First, identify what is actually moving across borders. A company should determine whether a project involves ordinary API access, software, model weights, source code, technical information, advanced computing hardware, cloud capacity, capital investment or some combination of those assets.

Second, identify every relevant party. Immediate customers may not tell the whole story. Headquarters, ultimate-parent relationships, intermediaries, cloud providers, beneficial ownership and end users can affect the analysis.

Third, determine where the technology will be located and used. The physical destination, remote access arrangements and location of computing infrastructure may all matter.

Fourth, establish the regulatory classification. Companies dealing with U.S.-origin technology should determine whether relevant hardware, software or technology is subject to the EAR and whether a license requirement or exception applies.

Fifth, examine the intended end use. BIS has specifically warned that advanced-computing items used for AI training can raise concerns where prohibited military-intelligence or weapons-related uses are involved. Its policy statement on advanced computing and AI model training provides additional guidance.

Sixth, assess investment exposure separately from export exposure. A transaction that does not involve an export can still require analysis under Treasury's outbound-investment rules.

The important point is that regulatory exposure depends on the technology, transaction, parties, ownership, destination and use. Political statements alone are not a substitute for determining which legal rules actually apply.

AI Governance Is Becoming a Geopolitical Risk Issue

AI governance is no longer only about privacy, bias, model transparency, safety and cybersecurity.

For internationally active organizations, governance can also include technology dependencies, export controls, geopolitical exposure, third-party ownership, cross-border restrictions and supply-chain resilience.

That does not mean geopolitical risk is itself a universal legal requirement.

It is better understood as an enterprise-risk and governance consideration that can help companies identify when an AI deployment, investment or partnership requires legal, security or compliance review.

For boards and AI governance teams, one practical consequence is that vendor due diligence may need to capture information beyond model performance and privacy practices. Supplier location, ownership, computing infrastructure, subcontractors and technology provenance can become relevant as well.

What Happens Next?

The clearest near-term development to monitor is the next U.S.-China AI dialogue.

Both governments say another exchange is expected in November 2026. The White House describes the mechanism as the U.S.-China Super Intelligence Dialogue, while China's Foreign Ministry calls it the China-U.S. AI Dialogue.

Businesses should also watch for future BIS changes affecting advanced computing and semiconductor controls, Treasury regulations implementing the COINS Act, Chinese policy responses, and any new bilateral arrangements addressing AI safety or technology transfer.

Any future U.S.-China corporate AI partnership would need to be evaluated on its actual structure. Shared research, a commercial API agreement, cloud access, an equity investment and a jointly owned frontier-model venture can raise very different policy and legal questions.

Trump's September 29 remarks draw a clear political line against the last of those when sensitive U.S. technological knowledge is at stake.

They do not, based on the evidence currently available, draw a line against every form of U.S.-China AI engagement.