US President Donald Trump has ruled out a joint venture with China to develop artificial intelligence, arguing that closer technological integration could expose sensitive American know-how to Washington’s main strategic competitor. Trump said the US is currently leading China in AI and questioned why Washington would share technology that could help narrow that gap.
The comments came shortly after Trump and Chinese President Xi Jinping agreed to establish a regular dialogue on the risks and benefits of what the administration calls “super intelligence.” The two countries also agreed to maintain a communication channel for AI-related incidents, showing that Washington and Beijing are willing to communicate about the technology even as they reject deeper integration of their AI development efforts.
Key takeaways
- Trump has ruled out a joint US-China venture to develop AI.
- He argued that cooperation could expose American AI secrets to China.
- Trump said the US is currently leading China in AI and wants to maintain that advantage.
- The US and China have nevertheless established a Super Intelligence Dialogue.
- The two countries also agreed to create a communication channel for AI-related incidents.
- The next US-China AI dialogue is expected by November 2026.
- Trump is prioritising technological leadership and voluntary safeguards rather than deeper international integration.
- The dispute highlights how AI is increasingly being treated as a strategic and national-security technology, rather than simply another commercial sector.
Trump rejects the idea of sharing AI technology with China
Trump’s position is based on a simple strategic calculation: if the US believes it has a significant lead in AI, it should not give a major competitor direct access to the technologies behind that lead.
Speaking at an event in Washington, Trump said that a joint venture would mean giving China access to secrets that the US possesses and China does not. He argued that there was little reason to share those capabilities while Washington believed it was ahead.
Trump also framed the competition in unusually stark terms.
He said the contest was essentially between the US and China, suggesting that whoever wins the race for advanced AI could gain a major strategic advantage.
That makes the administration’s position different from a blanket rejection of communication with Beijing.
Washington is willing to talk about AI risks.
It is not willing, at least under Trump’s current position, to combine the two countries’ frontier AI development efforts.
Why AI secrets are strategically important
The concern goes beyond individual AI applications.
The most valuable assets in frontier AI can include model architectures, model weights, training techniques, specialised datasets, inference systems, advanced chips and the infrastructure required to train increasingly capable models.
Model weights are particularly sensitive because they encode the parameters produced during training and can represent a substantial amount of the work required to develop a sophisticated model.
US lawmakers have increasingly focused on the possibility that adversaries could obtain access to sensitive model weights or other AI technology.
Representative Ro Khanna, a member of the House China Select Committee, recently asked major US AI companies including OpenAI, Anthropic, Google, Meta and SpaceXAI to provide information about attempted or successful unauthorised access to their model weights by China or other adversaries.
That concern provides important context for Trump’s position.
A joint AI development programme could theoretically create new channels through which technical information, research or infrastructure knowledge could move between the two countries.
Trump’s argument is that the potential strategic cost is too high.
The US and China are still talking about AI
Trump’s rejection of a joint venture does not mean Washington and Beijing have stopped discussing artificial intelligence.
In fact, the opposite is happening.
During Xi’s September 2026 visit, the two governments agreed to establish the US-China Super Intelligence Dialogue, according to the White House.
The mechanism is designed to allow the two countries to exchange views about the risks and benefits associated with advanced AI.
The governments also agreed to establish a bilateral communication channel for AI-related incidents.
The next exchange is expected by November 2026.
This creates an important distinction between AI governance dialogue and AI development cooperation.
The first involves talking about risks, incidents and potential safeguards.
The second could involve sharing technology, research or development capabilities.
Trump is supporting the first while rejecting the second.
US-China AI relationship
US–CHINA AI RELATIONSHIP
AI
│
┌──────────┴──────────┐
│ │
▼ ▼
COOPERATION COMPETITION
│ │
│ │
┌──────┴──────┐ ┌──────┴──────┐
▼ ▼ ▼ ▼
Risk Incident Frontier Technology
dialogue channel models race
│ │ │ │
└──────┬──────┘ └──────┬──────┘
│ │
▼ ▼
YES / OPEN NO / RESTRICTED
to dialogue joint development
Source: White House / Reuters reporting,
September 2026.
The emerging relationship can therefore be described as communication without technological integration.
Trump says the US wants to preserve its lead
Trump repeatedly described the US as being significantly ahead of China in AI.
That claim is central to his reasoning.
If Washington believes the US has a meaningful technological advantage, sharing frontier capabilities could reduce that advantage.
Trump said the US does not want to “mix it up too much” with China because cooperation could allow Beijing to gain knowledge that it currently lacks.
This approach places AI in the same broader strategic category as advanced semiconductors, computing infrastructure and other technologies with national-security implications.
The difference is that AI can spread through software, research and talent as well as physical infrastructure.
That makes controlling technological knowledge considerably more complicated.
AI is becoming part of US-China strategic competition
For much of the last decade, artificial intelligence was primarily discussed as a commercial technology.
Companies competed to build better models.
Governments discussed economic growth.
Investors focused on AI startups and data centres.
That picture has changed.
AI is now increasingly linked to national security, military capability, cybersecurity, economic productivity and geopolitical influence.
The US government’s restrictions on advanced semiconductor technology going to China have already demonstrated how Washington views computing infrastructure as strategically important.
Frontier AI adds another layer.
The country with stronger models may gain advantages in scientific research, software development, intelligence analysis, industrial automation and defence applications.
That is why Trump describes the competition as one in which there may be a major winner and loser rather than simply two companies competing for market share.
China wants a different approach to AI governance
Beijing has also argued for international cooperation on AI governance, although the two governments have significant differences over how that cooperation should work.
Before Xi’s visit, China’s foreign ministry warned against narratives based on threats and confrontation, arguing that such approaches could hinder international AI governance.
The contrast is becoming clear.
Washington is increasingly focused on protecting technological advantages.
Beijing is pushing for greater international participation in AI governance while simultaneously investing heavily in its domestic AI capabilities.
Both countries therefore have incentives to talk.
Neither has a strong incentive to give away its most valuable technological advantages.
Trump is not rejecting every form of AI cooperation
It would be misleading to describe Trump’s position as opposition to all US-China AI engagement.
The creation of the Super Intelligence Dialogue demonstrates otherwise.
The two governments have agreed to maintain a mechanism for discussing AI risks and incidents.
This could be useful even if the countries remain competitors.
For example, advanced AI systems could eventually be used in cyberattacks, influence operations or autonomous systems.
If an AI-related incident crosses national borders, Washington and Beijing could need a direct communication mechanism to determine what happened and prevent escalation.
That makes incident communication potentially valuable even without shared development.
Why an AI incident channel matters
AI systems are increasingly capable of acting through tools and software rather than merely generating text or images.
That raises the possibility of incidents involving:
- Cybersecurity systems
- Critical infrastructure
- Financial networks
- Autonomous software agents
- Military systems
- Cross-border information operations
- Sensitive government systems
If an AI system associated with one country accidentally or deliberately affects infrastructure in another country, policymakers may need to distinguish between an automated incident and an intentional attack.
A dedicated communication channel could reduce the risk of miscalculation.
The US and China have already recognised this problem by agreeing to establish a bilateral channel for AI-related incidents.
Trump is also resisting calls for tighter AI regulation
The rejection of a joint venture fits into Trump’s broader AI policy.
The president has repeatedly argued that excessive regulation could slow American AI development and weaken the country’s position relative to China.
In late September, Trump announced a voluntary AI safety agreement involving major technology companies including OpenAI, Anthropic, Google, Meta, xAI and Nvidia.
The agreement emphasises internal monitoring, external review and board-level oversight rather than a comprehensive government regulatory regime.
The administration is therefore pursuing two related objectives:
Move quickly enough to preserve US AI leadership.
Prevent the technology from creating unacceptable risks.
Trump’s preferred solution is largely voluntary safeguards and existing legal mechanisms rather than heavy new regulation.
The contradiction at the centre of the policy
There is an important tension in this approach.
The administration wants American companies to develop AI as quickly as possible.
At the same time, it wants to prevent China from obtaining access to the technologies being developed.
That means Washington has to encourage innovation while protecting increasingly valuable intellectual property.
The more capable AI becomes, the more difficult that balance could become.
Researchers move between companies.
Open-source models can distribute technology globally.
AI systems can be accessed through APIs.
Model outputs can potentially be used to improve competing systems.
And cybersecurity breaches can expose sensitive technical information.
The result is that protecting AI leadership is not simply a matter of banning technology exports.
It is also a question of protecting the entire AI development ecosystem.
The model-weight security problem
One of the most sensitive areas is model weights.
A frontier AI model can require enormous computing resources and engineering effort to train.
If an adversary obtains the final model weights, it may potentially gain access to a significant portion of the capabilities without having to repeat the entire training process.
That is why US policymakers are increasingly concerned about the security of model repositories and AI infrastructure.
Khanna’s recent request to major AI companies specifically asked about attempts to access sensitive model weights and the cybersecurity measures being used to protect them.
The issue demonstrates why Trump’s concern about “secrets” extends beyond traditional intellectual property.
The sensitive information could be the model itself.
China remains a major AI competitor
Trump’s decision is also shaped by the rapid development of China’s domestic AI industry.
Chinese companies are developing large language models, AI agents, robotics systems and specialised AI applications.
The US and China are therefore competing across multiple layers of the technology stack.
These include:
| Area | US–China competition |
|---|---|
| Frontier AI models | High |
| AI chips | High |
| Data-centre infrastructure | High |
| AI research | High |
| Robotics | Increasing |
| Autonomous systems | Increasing |
| AI cybersecurity | Increasing |
| AI governance | Significant |
| Safety dialogue | Limited but expanding |
The rivalry means that even cooperation designed around safety can become politically sensitive.
A technical standard agreed between Washington and Beijing could influence how companies around the world develop AI.
What Trump’s decision means for US AI companies
For American AI companies, the policy could reinforce the importance of keeping frontier technology within trusted jurisdictions.
Companies may face greater pressure to strengthen:
- Model-weight security
- Access controls
- Cybersecurity
- Employee security procedures
- Cloud infrastructure protection
- Export compliance
- Monitoring of foreign access
At the same time, US companies want access to international markets.
That creates a difficult commercial balance.
Restricting technology can protect strategic advantages, but excessive restrictions can also reduce a company’s global customer base and revenue opportunities.
What it means for China
For Chinese AI developers, Trump’s position reinforces the incentive to become technologically self-sufficient.
If access to American frontier technology remains restricted, Chinese companies have stronger reasons to develop domestic alternatives for chips, models, cloud infrastructure and AI tooling.
That could accelerate China’s push toward an independent AI technology stack.
Ironically, restrictions designed to preserve the US lead could therefore encourage China to reduce its dependence on American technology more quickly.
Whether that ultimately strengthens or weakens the US advantage will depend on how quickly each side can innovate.
The AI race is becoming a two-track system
The emerging landscape increasingly resembles two parallel systems.
The US-led ecosystem is anchored by companies such as OpenAI, Google, Anthropic, Meta, Nvidia and other technology firms.
China has its own rapidly developing ecosystem of AI companies and research institutions.
There will still be interaction between the two.
Researchers will publish papers.
Companies will compete globally.
AI models will be evaluated against each other.
Governments will communicate about risks.
But direct technological integration at the frontier is likely to remain limited.
Trump’s latest comments make that separation more explicit.
What happens next?
The immediate test will be whether the new US-China AI dialogue can produce meaningful communication without requiring the two countries to share frontier technology.
The November exchange could provide an early indication of whether the dialogue will focus primarily on incident management and risk communication or expand into broader questions around AI safety and governance.
The larger question is whether Washington and Beijing can develop rules for dealing with advanced AI while remaining intense technological competitors.
The Bigger Picture
Trump’s rejection of a joint US-China AI venture signals that Washington increasingly views frontier AI as a strategic asset rather than an ordinary commercial technology. The administration is willing to maintain dialogue with Beijing on AI risks and incidents, but it does not appear willing to combine the technological capabilities that underpin American AI leadership.
That creates a new model for international AI relations: cooperate on preventing dangerous incidents while competing aggressively on the technology itself. The approach may help the US protect sensitive AI capabilities, but it also risks deepening technological fragmentation and encouraging China to accelerate development of independent alternatives.
The biggest uncertainty is whether the US can maintain its lead while restricting technology transfers without slowing its own companies. AI moves unusually quickly, and the advantage Trump wants to preserve will depend not only on protecting existing secrets but also on continuing to produce better models, chips, infrastructure and applications faster than competitors.
Looking Ahead
The next US-China AI dialogue, expected by November 2026, will show whether the two governments can turn their new communication mechanism into something substantive. Discussions around AI incidents, safety and risk could become increasingly important as autonomous systems become more capable.
For now, however, Trump’s position is clear: Washington can talk to Beijing about AI risks, but it does not want to give China direct access to the technologies that could determine who leads the next generation of computing. That makes AI cooperation one of the clearest examples of how technology is becoming inseparable from US-China strategic competition.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
