Anthropic blocks AI models 'Mythos' and 'Fable' after US export curbs

AI company Anthropic has been compelled to suspend access to its "Mythos 5" and "Fable 5" language models for all non-U.S. citizens, including those residing within the United States. This decision stems from new U.S. export restrictions that necessitate the company's immediate compliance.

“

This directive means we must immediately deactivate Fable 5 and Mythos 5 for all our customers to ensure compliance with regulations.

"This directive means we must immediately deactivate Fable 5 and Mythos 5 for all our customers to ensure compliance with regulations," Anthropic stated in a notification. The company clarified that access to its other AI models remains unaffected by these new restrictions.

Anthropic has expressed disagreement with the U.S. government's assessment, which reportedly assumes these models could be circumvented through "jailbreaking." The company acknowledged that during a demonstration, the models exhibited a small number of "already known, minor vulnerabilities," but argued that similar weaknesses are present in other publicly available AI models. Anthropic considers the government's order unwarranted.

“

The access to all other Anthropic models remains unaffected.

While the company recognizes the government's authority to prohibit dangerous applications through a transparent and fair process, it contends that "this measure does not meet these principles." Anthropic maintains that the restrictions are disproportionate given the nature of the identified vulnerabilities.

“

This measure does not meet these principles.