Anthropic Disables Advanced AI Models Following US Government Order

The artificial intelligence company Anthropic has announced that it will stop access to its most advanced AI models, Fable 5 and Mythos 5. This decision follows a direct order from the United States government, which cited national security concerns. According to The Guardian Technology, the company will "abruptly disable" these models for all users to ensure it follows the new rules.

The government issued an export control directive—a set of rules that limits the movement of technology to other countries—requiring Anthropic to block access to these specific models for all foreign nationals. As reported by The Verge, this order applies to people both inside and outside the United States, including some of the company’s own employees. To meet these strict requirements, Anthropic decided that it had to cut off access to the models for all customers completely.

Anthropic stated that the government did not provide specific details about the security risks. Instead, the company claims that the evidence provided was only verbal and described a potential "jailbreak." In the world of technology, a jailbreak is a method used to bypass or break the safety rules that stop an AI from doing something it should not do. The government reportedly believes that these models could be used to find software vulnerabilities, which are weak points in computer systems that hackers might attack.

This situation marks a significant moment in the relationship between the US government and AI developers. For many years, US export controls have focused on physical items, such as the computer chips used to power AI. However, this new action shows that the government is now focusing on the software and the AI models themselves. The Verge noted that Anthropic believes the vulnerabilities discovered by the government are minor and that other AI models from rival companies have similar abilities.

There is clear disagreement between the two sides. Anthropic argued that it had worked closely with the US and UK governments on safety before launching its models. The company also pointed out that it has guardrails—safety features designed to prevent misuse—that stop the models from being used in risky areas like cyber-attacks. "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," the company said in a statement cited by The Guardian Technology.

Not everyone agrees with the company's position. Kirsten Davies, the chief information officer for the Pentagon, shared her thoughts on the social media platform X. She stated that the defense department supports prioritizing national security over business goals. "Some things are simply more important than revenue cycles, clickbait and pre-IPO valuation. America First. Always," Davies wrote.

This event comes at a difficult time for Anthropic, which recently filed for an IPO—a process where a private company offers its shares to the public for the first time. The company’s relationship with the government has been tense for some time. Earlier this year, Anthropic refused to let the US military use its AI for domestic surveillance or weapons systems. Because of this, the government placed Anthropic on a supply chain blacklist, which is a list of companies that are restricted from doing business with certain government partners.

Looking ahead, this situation highlights the growing tension between AI developers and government regulators. While Anthropic has previously called for more government oversight of AI, it argued that this specific action did not follow fair or fact-based rules. As the government continues to try to halt the AI capabilities of foreign adversaries, companies like Anthropic find themselves caught in the middle of a complex and changing legal landscape. For now, users of Fable 5 and Mythos 5 will have to wait to see if access is restored or if these models will remain unavailable.