Why is the U.S. ban on exporting Anthropic's top-tier AI models ineffective? Foreign media reveal through the "cryptocurrency war": history proves that restrictions are doomed to fail

The White House has recently issued an export ban on the well-known AI unicorn Anthropic, citing “national security,” restricting the overseas flow of its powerful Mythos and Fable models. However, in a column from international media, the piece directly points out that, looking back at historical failed regulatory cases targeting cryptocurrency technology (Crypto Wars) and spyware, trying to “contain” cutting-edge AI development through export bans is not only unrealistic, but may ultimately backfire on the United States’ own technological competitiveness.
(Background brief: Hurry up and lift the ban on Fable and Mythos! Anthropic promises to align more closely with the White House’s “reconciliation proposal”)
(Additional background: Elon Musk: Chinese large models will catch up to Anthropic’s Fable by Q1 2027! Zhipu Tang: It won’t take that long)

Table of Contents

Toggle

• The fuse that triggered the ban: South Korean telecoms and an Amazon report
• Lessons from history: Failures of crypto wars and spyware regulation
• Banning AI exports could backfire on U.S. competitiveness

Geopolitical power struggles in the field of artificial intelligence (AI) are heating up again. According to a recent special report published on June 19, 2026 by veteran journalist Lorenzo Franceschi-Bicchierai, last Friday the White House suddenly took a hard-line step, citing “national security,” and issued an emergency order requiring the well-known AI lab Anthropic to fully restrict exports of its leading AI models, Fable and Mythos, to any entities outside the United States—extending even to foreign nationals located inside the United States.

After receiving the notice, Anthropic was forced to cut off overseas access to these two major models within just 90 minutes, and it has been unable to supply them for over a week so far. This is the first time the U.S. government is truly testing, in practical terms, the idea of using “export controls” to contain frontier AI technology. The outcome will not only affect Anthropic’s international footprint, but will also become a benchmark for global AI regulation.

The fuse that triggered the ban: South Korean telecoms and an Amazon report

What exactly led the U.S. government to take such a drastic action to crack down and block it? Reports say the trigger mainly came from two major events. First, Anthropic previously granted a South Korean telecom company (reportedly SK Telecom) access to Mythos through a limited partner program. However, U.S. officials suspected that the company might have some kind of connection to China; even though the telecom company issued a firm denial, it still touched a sensitive nerve in Washington.

Second, Amazon’s CEO Andy Jassy reported to the U.S. government that internal researchers had found a vulnerability to bypass the security safeguards of the Fable 5 model. Even though Anthropic emphasized that this was a narrow, patched issue and not a full “jailbreak,” the Department of Commerce still decided to impose controls immediately. In fact, since Mythos was launched in April this year, Anthropic has positioned it as a “doomsday-level cyber weapon.” To guard against cyberattacks, it previously opened access to only about 150 strictly vetted organizations—intended to “let defenders get the weapon first.”

Lessons from history: Failures of crypto wars and spyware regulation

However, is relying on “export controls” to contain emerging dual-use technologies really effective? The author of the article says bluntly that historical experience provides an answer in the negative.

The most famous example is the “Crypto Wars” of the 1990s. At the time, the U.S. government viewed encryption technologies such as PGP as weapons, worrying that intelligence agencies would be unable to intercept and monitor, and even launched a criminal investigation into the developer Phil Zimmermann. As a countermeasure, Zimmermann chose to print the source code into a physical book and publish it. In the end, the investigation was forced to be dropped, and it instead helped usher in the era of end-to-end encryption that is now widely used by Signal, WhatsApp, and others. This is regarded as the most notorious failure case in the history of export controls.

Another example is spyware in the 2010s. Although multiple Western countries classified surveillance software as dual-use technology through the Wassenaar Arrangement and required export licenses, the effectiveness was undermined because some countries did not participate (such as Israel), some participating countries enforced the rules loosely (such as Italy, which once allowed exports to oppressive regimes), or companies directly relocated their headquarters to countries with more permissive controls (such as Saudi Arabia). As a result, the regulation became effectively a nullity.

Banning AI exports could backfire on U.S. competitiveness

In the concluding section, the special report warns that the current deadlock between Anthropic and the Trump (Donald Trump) administration remains ongoing. The U.S. government’s options in the future appear to fall into a dilemma: if it concedes and lifts the restrictions, it amounts to acknowledging that AI labs of competing countries such as China will, regardless, catch up; but if it insists that any future AI model exports from the United States must go through strict government approval, it will significantly raise compliance costs for companies and severely impair the international market share and profitability of U.S. AI companies.

Based on the history of repeated battles and repeated failures in regulating crypto technology and spyware, harsh export controls are probably never the right solution to prevent malicious actors from abusing powerful technologies.