When nations believe they possess a strategic technological advantage, they face a choice. They can protect that advantage by restricting access to it. Or they can spread it so widely that the rest of the world gets hooked on their technology.
China appears to be choosing the first path.
Recent reports indicate that Beijing is considering tighter controls on the export of its most advanced artificial intelligence models, treating frontier AI as a national security asset available to foreign users only under restrictive conditions. If true, this suggests that China’s leadership has concluded that the country’s long-term success depends on jealously guarding its best technologies.
The United States should take the opposite approach.
America has repeatedly become the world’s technological leader not by keeping its innovations at home, but by making them indispensable abroad. American operating systems, cloud computing, software platforms, financial services, entertainment, and internet companies became global standards because businesses and consumers everywhere wanted to use them. This meant enormous profits for U.S. companies—and expanded influence for the U.S. It also ensured that America remained the top destination for global talent, and set the stage for future innovation that would further U.S. dominance.
We have the opportunity to do this again with AI. The United States leads the world in AI innovation and investment. Our frontier models have always been best-in-class. But the gap is narrowing.
China’s AI firms have dramatically improved their capabilities over the past two years. In June, a Chinese developer released a new AI model that may be able to perform on par with Anthropic’s cutting-edge Mythos on cybersecurity, which could mean that the Chinese government has the ability to bypass or disarm many of our existing cyber defenses. Worse, this model was reportedly trained on indigenously produced chips, suggesting that the problem cannot be solved by slapping export controls on advanced U.S. semiconductors.
How should Washington respond?
Some policymakers are looking at restricting foreign entities’ access to American AI systems. This was the administration’s strategy with Anthropic, which just saw access restrictions on some of its most advanced AI models lifted. In Congress, proposals like the Remote Access Security Act would extend export controls beyond advanced computer chips to include cloud-based AI services.
There are circumstances in which this makes sense—when AI capabilities have clear military applications, for example, or when adversarial governments are involved. But we should be very careful to distinguish between the denial of advanced capabilities to our strategic rivals and the throttling of access for everyone else. If the U.S. government blocks legitimate international customers from doing business with American companies, the effect will not be to eliminate foreign demand. It will be to redirect that demand toward competitors.
History offers a cautionary lesson. Export restrictions can work, for a while. But they create strong incentives for alternative sources of supply—and once those emerge, they tend to stick around. Competitors become more self-sufficient, U.S. companies lose markets, and American technological influence declines.
AI is especially prone to this dynamic, both because of the speed with which the technology changes and the degree to which success depends on scale. The more users a given model has, the more data and feedback its developers receive to improve it, and the more complementary software and expertise accumulate around it. Those network effects make leading AI systems stronger over time. Restricting exports would thus mean foregoing those advantages while incentivizing the emergence of viable, non-American AI ecosystems.
Instead of hunkering down, we should look at Beijing’s misstep as Washington’s opportunity. China’s desire to control and hoard its capabilities will backfire, lessening the country’s influence and ultimately constraining its technological, economic, and military ambitions.
The United States should therefore pursue a different strategy. Rather than limiting access to the best our tech sector has to offer, we should seek to make American AI the preferred platform for businesses, researchers, and governments around the world. Countries that build their economies around U.S. technology are more likely to adopt our technical standards, buy our products, collaborate with our researchers, and remain engaged with our industry. By addicting global consumers to the U.S. tech stack, we will maximize American prosperity and American influence for the benefit of all.
The contest for primacy in the 21st century will not be won by denying advanced AI to others. It will be won by making American technology so ubiquitous and essential that the world cannot live without it.
Read the full article here
