US Orders Anthropic to Halt Powerful AI Access Citing 'Super Weapon' & National Security Fears

Top Summary

• What happened: The US government ordered Anthropic to immediately revoke access for foreign nationals to its powerful AI models, Claude Fable 5 and Claude Mythos 5, citing national security concerns. To comply, Anthropic suspended access for all users globally.
• Why it matters: These AI models, particularly Mythos, were previously described as 'super weapons' capable of identifying critical system flaws and potentially building bioweapons, raising significant concerns about misuse and national security.
• What changes: Access to Anthropic's most capable publicly available AI models, Fable 5 and Mythos 5, is now suspended for all customers.
• Who is affected: All customers and foreign nationals who previously had access to Anthropic's advanced AI models, Fable 5 and Mythos 5, are directly impacted by this suspension.

Anthropic's AI Models Deemed 'Super Weapons'

Anthropic, a leading AI company, is facing a major setback after the US government intervened over its powerful artificial intelligence models. The company's underlying model, Mythos, which Fable 5 is built upon, was reportedly warned by testers as a 'super weapon,' requiring a 'gun license' for its use.

Mythos was previewed by Anthropic in April but kept from mass release due to its immense power. Concerns emerged that malicious actors could exploit it to hack critical infrastructure like banking systems or to develop bioweapons.

According to the company, Mythos successfully identified flaws in every major operating system and web browser it tested, some vulnerabilities reportedly lying undetected for decades. To mitigate risks, Anthropic launched Project Glasswing, a controlled program sharing Mythos with around 50 vetted organizations, including Google, Apple, Amazon, Microsoft, and CrowdStrike, for defensive cybersecurity work.

Government Intervention and Global Suspension

Last Tuesday, Anthropic released a version of Mythos with built-in guardrails, known as Fable 5, designed to block responses in high-risk areas such as cybersecurity and biology. This version was ranked as the most capable publicly available AI model on benchmark tests by Vals AI.

However, the US government intervened on Friday, issuing an 'export control directive' that ordered Anthropic to immediately revoke access for foreign nationals to Claude Fable 5 and Claude Mythos 5. The directive applied to all foreign nationals, both inside and outside the US, citing national security concerns.

In response, Anthropic suspended access for all its customers globally to comply with the government order. The company stated that this action was necessary to ensure full compliance with the directive.

"Misunderstanding" Amidst Jailbreak Concerns

"The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," Anthropic said in a lengthy statement. The company believes "this is a misunderstanding" and is actively working to restore access "as soon as possible."

The government's directive was prompted by concerns of a potential "jailbreak." A jailbreak refers to techniques used to bypass an AI model's built-in safety restrictions, coaxing it to provide responses it would normally refuse.

Anthropic stated that the government provided only verbal evidence of a "potential narrow, non-universal jailbreak," indicating the issue might not be widespread but was serious enough to trigger the drastic action.

What to Watch Next

Anthropic is engaged in active discussions with the US government to clarify the situation and potentially restore access to its powerful AI models. Future developments will depend on these negotiations and any further directives issued by authorities regarding AI safety and national security protocols for advanced technologies.