Initially, AI’s critics insisted that artificial intelligence was just another software product. AI was presented as a huge commercial opportunity, sure. It was presented as a tool through which humans could enhance their lives, but ultimately it was still understood as a statistical program that knew how to spell. Thanks to the Trump administration’s Anthropic export ban, that illusion is dead.
The more powerful the technology becomes, the more determined governments are to control who can access it
The United States government ordered Anthropic to suspend access for non-US persons to Fable and Mythos 5, its most advanced models, after officials raised national-security concerns. Whatever one thinks of the decision itself, its significance is hard to overstate. Washington has effectively declared that frontier models belong in the same regulatory category as weapons and other strategic technology.
This shouldn’t come as a surprise. Throughout history, technology capable of altering the balance of power has quickly ceased to be treated as an ordinary commercial good. The United States has a history of preventing strategically important technology from reaching potential adversaries. The Anthropic export restriction follows semiconductor export controls, and American intelligence agencies have gone so far as to acquire companies like Crypto AG in order to maintain technological advantages over rivals.
The language surrounding AI has not caught up with reality. Politicians still talk about innovation and regulation. Executives discuss the future of work. Investors talk about valuations and revenue. Critics talk about all of those things. Governments, meanwhile, are treating this generation of large language models as a geopolitical asset rather than a consumer product. A software company is typically free to sell its products to any willing buyer. A manufacturer of strategic assets, like fighter jets, is not. Strategic assets are restricted and controlled, and their distribution is discussed like a matter of state policy rather than of market forces.
The Anthropic saga suggests that governments have begun regarding AI as a technology too powerful to leave to the market. If AI systems can identify vulnerabilities, accelerate military planning, or help expedite scientific discoveries, then access to those systems becomes a question of national security.
For decades, globalization encouraged the belief that technological progress would dissolve national boundaries and bind humanity ever closer. AI appears to be producing the opposite effect. The more powerful the technology becomes, the more determined governments are to control who can access it.
The result might be an AI governance regime that looks less like the internet and more like nuclear proliferation. Access to the most advanced systems will not be determined solely by who can afford to run them, but by states, alliances, export controls and national interests. Fable and Mythos 5 are likely to eventually return. The broader trend, however, is unlikely to reverse. The question no longer is who will build the most advanced models, but who will be permitted to use them.
Comments