White House Compels Anthropic to Shut Down Two AI Models Over Security Risks

Fable 5 and Mythos 5 Models Taken Offline

According to НВ — Техно: On June 15, 2023, Anthropic deactivated its Fable 5 and Mythos 5 models at the White House’s request. The move came amid national security concerns and violations of U.S. export restrictions, which bar non-U.S. citizens from accessing these systems. The shutdown followed the Trump administration’s dissatisfaction with how Anthropic responded to safety complaints.

Anthropic, co-founded by Dario Amodei, had tried to negotiate an agreement with the White House as early as Sunday, but those efforts failed. U.S. authorities then imposed export controls prohibiting access to cutting-edge models-including Fable 5 and Mythos 5-for anyone without American citizenship. Both models belong to the Mythos class, which had been marketed as safer iterations of the underlying technology.

Safety Concerns Mount

Paid Claude users gained access to Fable 5 and Mythos 5 on Tuesday, but by Friday evening the models were pulled. The shutdown may have been triggered by reports of potential security bypasses. On Thursday night and Friday morning, Amazon-along with five other firms-alerted the White House to possible ways around the models’ safeguards.

This alarmed the White House, as intelligence suggested a group linked to China may have obtained access to a Mythos-class model.

Although Anthropic does not officially grant model access to users in China, White House officials told the company on Friday afternoon that continued operation of Fable 5 and Mythos 5 posed a national security threat. Anthropic was given 90 minutes to disable the models, after which an official directive citing export restrictions was sent. This incident underscores the critical need to comply with export regulations and maintain security in today’s technology landscape.

This case highlights growing U.S. government concerns about AI safety, especially amid global competition and risks tied to advanced technology access. The shutdown of Anthropic’s models may set a precedent for other AI companies and emphasizes the importance of adhering to regulatory requirements to prevent potential national security threats.

As concerns surrounding AI security continue to escalate, the recent actions taken by the White House highlight the ongoing challenges faced by companies like Anthropic. For a deeper understanding of the implications of strict restrictions on AI models, you can read more about the intense backlash against Anthropic's new model here.