SAN FRANCISCO — Over 100 cybersecurity professionals and executives are urging the Trump administration to reverse a government directive that blocks foreign nationals from accessing Anthropic’s most cutting-edge artificial intelligence models, warning that the restriction could end up benefiting America’s rivals more than protecting the country.
Anthropic announced Friday that it had taken two of its newest AI models — known as Fable 5 and Mythos 5 — offline in order to comply with the government’s order. The company made clear it did not believe the action was justified by the security concern the government had raised.
The San Francisco-based company had previously limited access to some of its most advanced technology to a select group of customers, citing the models’ ability to outperform human cybersecurity professionals in identifying and exploiting software weaknesses. Anthropic had held prior discussions with the White House regarding the capabilities of these newer models.
In a letter released Sunday, more than 100 cybersecurity professionals and leaders from companies including Adobe and Nvidia called on the federal government to lift the export control directives targeting Anthropic’s models. They also urged the administration to “commit to an open, scientific and transparent process of handling AI risk assessments in the future.”
The letter acknowledged that Anthropic’s Mythos models are “quite good” at uncovering software vulnerabilities and turning them into exploits, but argued they are “not uniquely good at these tasks.” Many of those who signed the letter said they regularly rely on other foundation and open-source models for security audits and training purposes.
According to the letter, stripping away top-tier cyber defense tools “without a good reason” is dangerous at a time when America’s adversaries are rapidly closing the gap. The letter noted that China’s AI models are “only months behind the best American models” and suggested China’s government likely has access to advanced capabilities that have not been made publicly available.
The export controls represent the most significant step the U.S. government has taken to date in limiting access to highly advanced AI systems. Anthropic had released Fable broadly just last week — a scaled-back version of the more powerful Mythos model, which the company had already been keeping under tight wraps due to cybersecurity concerns.
The Commerce Department had not responded to a request for comment as of Monday.
The directive came just 10 days after President Donald Trump signed an executive order creating a framework allowing the federal government to evaluate the national security implications of the most advanced AI systems for up to one month before they are released to the public. The order specified that participation by AI developers would be voluntary.
Relations between the Trump administration and Anthropic have been strained. The company has pushed for guardrails on AI development to reduce risks and strengthen both economic and national security benefits for the United States.
The tension escalated following a contract dispute with the Pentagon, after which Defense Secretary Pete Hegseth moved to label Anthropic a supply chain risk — an extraordinary step against a domestic company that Anthropic has since challenged in two separate federal courts. The company said it sought guarantees that the Pentagon would not deploy its technology in fully autonomous weapons systems or for surveilling American citizens. Hegseth countered that the company must permit any use the Pentagon considered lawful.
