Three Months After U.S. Defense Department Labeled Anthropic a 'Supply Chain Risk'

Government Bans Foreign Access to Phaible 5 Citing National Security

Restrictions Extend to Allies and Even Company Employees

'The world's leading artificial intelligence (AI) startup' Anthropic has come into direct conflict with the Donald Trump administration. This comes about three months after the U.S. Department of Defense designated Anthropic as a 'supply chain risk.' The U.S. government has banned foreign access to Anthropic's latest AI models—Phaible 5 and Mythos 5—citing national security concerns.


Although the U.S. government states that such measures will not be broadly applied across the information and IT sectors, the prevailing view in the industry is that AI models themselves are increasingly being treated as national security assets. Some experts have even described the move as an 'AI version' of the McMahon Act, which was enacted after World War II to prevent the sharing of nuclear technology.


Anthropic: "Even Foreign Employees Banned from Access"

Dario Amodei, CEO of Anthropic, spoke at an AI event in New Delhi, India, on February 19 (local time). Photo by Reuters Yonhap News

Dario Amodei, CEO of Anthropic, spoke at an AI event in New Delhi, India, on February 19 (local time). Photo by Reuters Yonhap News

View original image

Howard Lutnick, U.S. Secretary of Commerce, sent a letter directly to CEO Dario Amodei notifying him of these restrictions, according to reports from The Wall Street Journal (WSJ) and other outlets on June 13 (local time). Officials judged that blocking foreign governments, corporations, and individuals from accessing the relevant AI models was the most appropriate course of action.


Anthropic immediately suspended services for both models. The restrictions did not target specific countries like Russia or China; instead, all 'foreign nationals' were included, meaning that even 'Five Eyes' allies such as the UK, Canada, Australia, and New Zealand—as well as the UK AI Safety Institute—were completely blocked from access. The ban also extends to all foreign nationals both inside and outside the U.S., including foreign employees of the company itself.


The government raised concerns about the potential for 'jailbreaks'—methods to bypass safety mechanisms. Andy Jassy, CEO of Amazon.com, alerted the U.S. government to the possibility of a jailbreak in Phaible 5, prompting the White House to demand that Anthropic either fix the issue or suspend its service. Anthropic, in response, suspended the two models. However, the company countered that the vulnerability is one commonly found in other AI models and does not constitute a complete jailbreak.


The government and Anthropic have offered conflicting explanations for this extreme decision. David Sacks, a White House adviser, posted on his social network X (formerly Twitter) that "the government demanded Dario fix the jailbreak issue in Phaible or suspend service," and "Dario refused." In contrast, Anthropic insisted that it simply received instructions to block foreign access for national security reasons.


Some observers see this not just as a pure national security measure but also as a political clash between the Trump administration and Anthropic, who have been fiercely at odds over the weaponization of AI. In early March, President Trump banned all federal agencies from using Anthropic's AI model 'Claude' and designated Anthropic as a 'supply chain risk company.' Anthropic responded by filing suit against the U.S. government. President Trump also criticized Anthropic on his social media platform Truth Social, calling it a "radically leftist 'woke' company."


Kate Korn, Deputy Director of Economic Studies at the Center for Strategic and International Studies (CSIS) and a former Commerce Department official, told the WSJ, "While I understand the security concerns, it appears that the White House's animosity toward Anthropic may have influenced this decision."


U.S. Moves to Manage 'Advanced AI Technology' as a National Strategic Asset?

U.S. President Donald Trump. Photo by AFP News Agency

U.S. President Donald Trump. Photo by AFP News Agency

View original image

As various analyses emerge regarding the Trump administration's actions, there are predictions that this AI technology ban will become a 'geopolitical watershed.' The Economist noted that the situation is similar to when the U.S. government restricted the overseas use of public key encryption—digital communications security technology—in the 1970s-1990s, treating cryptography as a weapon. Ultimately, civil liberties advocates prevailed, and most encryption technology was permitted for use, sale, and export. The magazine also pointed out that the current situation is reminiscent of the McMahon Act during World War II, which blocked the sharing of nuclear technology.


Industry insiders now believe the government has started to manage 'advanced AI technology' as a national strategic asset. In fact, on June 2, the Trump administration signed an executive order requiring AI companies to implement a self-regulation framework that grants the government pre-release access to cutting-edge models. However, according to IT outlet The Information, citing U.S. officials, the Trump administration does not plan to extend these foreign access restrictions to other AI companies such as OpenAI.


Time magazine noted that while the U.S. government has used export controls to restrict semiconductor chip exports in the past, it has never before targeted AI models themselves. The report highlighted that this order demonstrates both the U.S. government's strengthening efforts to prevent foreign adversaries from using U.S.-developed AI technology, and the fact that the U.S. increasingly regards AI models as national security assets.


Key Allies Also Alarmed

Reuters Yonhap News

Reuters Yonhap News

View original image

The effects have been just as urgent for U.S. allies abroad. The Economist reported that the shock spread among allied nations when even the 'Five Eyes' intelligence partners lost access to Mythos. The UK AI Safety Institute was also blocked without exception. Some analysts suggest that this situation could prompt countries to reduce their reliance on U.S. AI and move to secure their own 'AI sovereignty.' However, it is reportedly the case that negotiations are underway to restore access for intelligence agencies.


Kaniska Narayan, the UK's Minister for AI and Undersecretary for Science, Innovation, and Technology, emphasized on X that "the most important lesson is that access to AI capabilities is essential when discussing the future of national security and technological sovereignty," stressing the need for increased investment. Canadian Prime Minister Mark Carney also characterized the incident not as a simple service suspension, but as an issue of AI sovereignty. He told reporters recently, "What we are experiencing now is what can happen when there is excessive reliance on a single AI model."



Meanwhile, Bloomberg reported that this ban has become a major setback for Anthropic, which is planning an initial public offering (IPO) later this year. Anthropic has selected Morgan Stanley and Goldman Sachs as lead underwriters. According to sources cited by the media outlet, the company is targeting an IPO in October this year. The company's valuation was recently estimated at $965 billion (about 1.455 trillion won) during its latest funding round. With this, Anthropic overtook its rival OpenAI to become the world's most valuable AI startup.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing