The United States government has ordered AI company Anthropic to suspend access to its most advanced artificial intelligence models for foreign nationals, citing national security concerns over the technology’s potential misuse.
In response to the directive, Anthropic announced on Friday that it would disable access to its recently launched Fable 5 and Mythos 5 models for all users globally to ensure compliance, describing the move as abrupt and unprecedented.
According to the company, the U.S. Commerce Department issued an export control directive based on concerns that the models could potentially be “jailbroken” to reveal information that might assist in identifying software vulnerabilities. Anthropic said it was not provided with detailed evidence supporting the government’s concerns, adding that officials had only cited what it described as a “narrow, non-universal” safeguard bypass.
“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.
The directive marks a significant expansion of U.S. efforts to control advanced AI capabilities. Previous export restrictions have largely focused on restricting access to advanced semiconductors and AI infrastructure rather than limiting access to AI models themselves.
The dispute comes amid broader tensions between Anthropic and the U.S. government. Earlier this year, the company reportedly declined to allow the military to use its AI systems for domestic surveillance and fully autonomous weapons programs, a decision that contributed to a deterioration in relations with federal authorities.
Anthropic recently introduced Fable 5, part of a new generation of “Mythos-class” AI models. The company said the models include safeguards designed to prevent misuse in areas such as cybersecurity, though some users had criticized those restrictions as overly broad.
Security experts have warned that highly capable AI systems could potentially accelerate sophisticated cyberattacks if deployed without adequate safeguards, particularly against critical sectors such as banking and financial services.
The company said it had worked with government agencies and external experts on safety measures before releasing the models and argued that competing AI systems demonstrate similar capabilities in identifying minor software vulnerabilities.
Anthropic also stated that the directive could have far-reaching implications for the wider AI industry.
“If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” the company said.
The company added that it is seeking clarification from regulators and hopes to restore access to the models as soon as possible.
Amazon CEO Reportedly Raised Concerns
Meanwhile, a report by the The Wall Street Journal claimed that the government’s decision was triggered after Andy Jassy, chief executive of Amazon, raised concerns directly with senior U.S. officials, including Scott Bessent.
According to the report, Amazon researchers were able to prompt Fable 5 into generating information that could potentially aid cyberattacks despite the model’s built-in restrictions. The findings reportedly prompted White House discussions and independent testing by security researchers.
Officials subsequently asked Anthropic to address the alleged vulnerabilities or withdraw the model. After reviewing the issue, the administration reportedly concluded that restricting access for foreign governments, businesses, and individuals was the most direct way to mitigate potential risks.
The report also stated that Donald Trump approved the action despite concerns within parts of the administration that such restrictions could slow AI innovation.
Anthropic has maintained that it is willing to engage with regulators and believes the situation stems from a misunderstanding. However, the company acknowledged that the directive could affect its own workforce, noting that several employees involved in developing its latest models were born outside the United States.
The company recently filed confidentially for an initial public offering, placing it in direct competition with rival OpenAI in the race to access public markets. The latest regulatory action is expected to intensify debates over AI safety, national security, and government oversight of increasingly powerful AI systems.
