AI and National Security: US Bans Export of Anthropic Models

The US government has ordered Anthropic to restrict the export of its most powerful AI models — Fable and Mythos — due to national security concerns. This ban applies not only to foreign states but also to foreign citizens within the US. Following this unexpected decision, the AI giant completely removed both models from use, and they have been unavailable to anyone for a week. This is reported by Techcrunch.com report says.

This event is a unique test for the US government. Whether the Washington administration can curb advanced AI technologies through export control mechanisms remains to be seen. Previous similar attempts related to encryption technologies and spyware did not yield the expected results. According to ixbt.com, this conflict could set new rules that not only Anthropic but all other AI laboratories must follow.

Why were Mythos and Fable deemed dangerous?

When Anthropic introduced the Mythos model in April, it described it as a powerful tool that could cause significant devastation in cyberspace. Therefore, even before the ban, access to this model was granted only to 150 vetted companies and government organizations. The goal was to help defenders strengthen their software before malicious actors could gain such capabilities.

It is said that two main events triggered the activation of the ban. First, Anthropic allowed South Korea's SK Telecom operator to use the Mythos model. US officials suspect the company's ties to China, although SK Telecom has denied these claims. Second, Amazon CEO Andy Jassy warned the government that a way to bypass the protection systems in the Fable 5 model had been found. While Anthropic called this error minor, the US Department of Commerce immediately issued an export control directive.

Historical Experience: Does Export Control Work?

Governments' attempts to limit the spread of cyber technologies are not new, but history shows few successes in this regard. In the 1990s, the US government viewed the PGP (Pretty Good Privacy) encryption software as a weapon and opened a criminal case against its creator, Phil Zimmermann. However, Zimmermann published the program code in book form, placing it under freedom of speech protection, which paved the way for the encryption technologies used today in messengers like Signal and WhatsApp.

Additionally, in the early 2010s, after it became known that Western-made spyware was used against dissidents in the Middle East, the Wassenaar Arrangement was expanded internationally. This agreement aims to restrict the export of technologies that can be used for both civilian and military purposes. The situation with Anthropic is the next important step in showing how effective these methods are in the age of AI.