U.S. President Donald Trump’s new executive order on artificial intelligence (AI) signals a sharper strategic turn in Washington’s technology policy. The order brings the U.S. national security state closer to the companies developing AI models. Its core purpose is to preserve U.S. innovation while ensuring that the most powerful AI capabilities support U.S. cyber defense, critical infrastructure and strategic competition with China. The result may be a more divided global AI order, with Chinese models facing growing scrutiny over security, data and political alignment.
From Risk Management to Strategic Competition
Trump’s order “promoting advanced artificial intelligence innovation and security,” released on June 2, appears at first glance to be a regulatory document. It establishes a framework for identifying protected frontier models, encourages voluntary cooperation between companies and federal agencies, and strengthens the use of AI in cyber defense.
But the order is more than an attempt to manage the risks of new technology. It is a statement about how Washington now understands AI in the context of China-U.S. competition. Advanced AI models are no longer treated simply as commercial products or tools for productivity. They are increasingly regarded as strategic assets linked to national power.
This marks a clear shift in the direction of U.S. AI governance. During the Biden administration, AI policy placed considerable emphasis on privacy, discrimination, consumer protection, misinformation and model safety. National security concerns were present, but they sat alongside a wider social risk agenda.
Trump’s new order gives priority to a different concern. It places innovation, cyber security and geopolitical competition at the center of AI policy. The logic is close to Trump’s first term approach in the 2019 American AI Initiative. AI leadership depends on research, capital, computing power, talent, and standards. Excessive regulation, in this view, risks weakening the very ecosystem that gives the U.S. its advantage.
Frontier AI Models as National Security Assets
The order therefore accepts two realities. The first is that frontier AI models now carry national security consequences. This is why the National Security Agency and the Cybersecurity and Infrastructure Security Agency have such prominent roles in the order. The main concern is not general AI ethics. It is whether the most advanced models can improve cyberattack capabilities, expose critical infrastructure to new risks, or be used by hostile actors.
The order asks U.S. agencies to develop a classified evaluation system for advanced AI models with high level cyber capabilities. It also allows companies to give the government early access to models before wider release. This is presented as a voluntary process, not a compulsory approval mechanism.
This design is important. Washington wants earlier insight into the capabilities of frontier models, but it does not want to create a formal licensing regime. The administration appears to believe that formal approval would damage the speed of AI development. In a sector where progress can occur in months, delay itself becomes a strategic cost.
This is the second reality underpinning the order: that an arduous approval system could slow American innovation and strengthen China’s relative position.
This is also where the U.S. approach differs from the European Union’s AI governance model. The EU has placed greater weight on risk classification and legal compliance. The Trump administration has chosen a more flexible framework. It seeks closer government access to powerful models while avoiding rules that could slow private sector development.
The result is a distinctive form of state industry cooperation. AI companies are not being placed under direct state control, but the government is building channels to evaluate, influence and use their most advanced systems. Companies that cooperate may gain access to federal procurement, defence networks and trU.S.ted partner arrangements.
Looming Competition With China’s AI Industry
This could help Washington turn commercial innovation into national capability and also reshape the AI market. A voluntary assessment system may become a practical requirement if investors, corporate customers, and government agencies begin to treat it as a mark of trust. Large firms such as OpenAI, Google, and Anthropic are better placed to manage such relationships. Smaller firms and open-source developers may find it harder to operate in a market where government validation becomes a commercial advantage.
The China implications are significant. The order does not directly name Chinese companies, but its strategic meaning is clear. AI is moving into the same policy category as semiconductors, telecommunications, and critical minerals. It is a technology that Washington sees as too important to be left to open market competition alone.
Chinese AI firms have already entered parts of the U.S. digital ecosystem through open-source models, developer platforms, and consumer applications. Some Chinese models are used by developers outside China. Chinese AI applications in social companionship, image generation, and other consumer services have also reached users in the U.S..
These companies are unlikely to benefit from the voluntary testing framework in the same way as U.S. companies. They are not natural candidates for U.S. government contracts. Nor is Washington likely to treat them as trusted partners in cyber defence or critical infrastructure.
But the indirect effect could be substantial. If Chinese companies do not submit models for U.S. security testing, their refusal may be interpreted by critics in Washington as evidence of risk. They may be accused of hiding backdoors, transferring data to China, or aligning with Chinese state objectives. If they do participate, they may face difficult questions about what information must be shared with U.S. national security agencies and whether such sharing would conflict with Chinese laws on data security, state secrets, or technology export control.
A Two Track System in a More Divided World?
This points toward a likely two track system. U.S. and allied models may enter a trusted ecosystem built around government testing, procurement, and security cooperation. Chinese models may face a separate track of scrutiny focused on data, influence, censorship and political alignment.
Content control may become especially important. Chinese generative AI services are required to comply with domestic rules that include political and ideological content standards. U.S. officials could use this as evidence that Chinese models are not politically neutral. Testing a model on sensitive political questions may become a way of framing it as subject to Chinese state influence, even without proving direct data theft or cyber risk.
The longer term consequence is a more divided global AI order. China-U.S. competition will not only be about who builds the best models. It will also be about who sets standards, who is trusted, who gains market access and whose models are treated as safe for use in critical systems.
Trump’s AI order does not slow AI development. Its purpose is to make AI development serve American strategic power. That is why the order matters beyond domestic U.S. regulation. It shows that Washington now sees frontier AI as central to the next phase of technology competition with China.