U.S. Top AI Firm Anthropic Halts Access to Its Most Powerful AI Systems Upon Government Request

Leading American artificial intelligence company Anthropic cut off access to its most powerful models on Friday, complying with a government directive just three days after the models were released. The unprecedented move by Washington was justified on grounds of national security risk.

Anthropic announced the suspension of two new models—Mythos 5 Lite and Fable 5, a streamlined version for general users—to adhere to an executive order issued by the government that evening.

According to a press release published by Anthropic, Washington cited export control regulations, demanding the termination of "any foreign citizen's access" to these models, whether located within or outside the United States—including the company’s own “foreign employees.”

Due to the inability to accurately identify users and existing tensions with the Trump administration, Anthropic stated it had no choice but to “immediately disable” both models, suspending services to all customers entirely.

According to U.S. media Axios, the directive was issued by U.S. Commerce Secretary Howard Lutnick.

It is reported that he took action after learning a company had successfully bypassed the security safeguards of these Anthropic models. These models are known for their unprecedented speed and precision in detecting and exploiting cybersecurity vulnerabilities.

AFP has reached out to the U.S. Department of Commerce for comment but has not received a response.

In a statement, Anthropic said: “We do not agree that the mere discovery of a potential circumvention method justifies recalling a commercial model already deployed to hundreds of millions of users.” The company hopes this incident will be seen as a “misunderstanding.”

Anthropic added: “If this standard were applied across the entire industry, we believe it would effectively bring all new deployments of cutting-edge AI models to a standstill.”

Currently, the company stands at the forefront of global competition with OpenAI, Google, and China’s DeepSeek.

Anthropic stated that it is working toward “restoring access” to these latest models as soon as possible, while other models remain operational.

Source: rfi

Original: toutiao.com/article/1867900527759436/

Disclaimer: This article represents the personal views of the author