• Wion
  • /World
  • /US locks down Anthropic's most powerful AI after alarming jailbreak warning

US locks down Anthropic's most powerful AI after alarming jailbreak warning

US locks down Anthropic's most powerful AI after alarming jailbreak warning

This photograph shows the logo of the US artificial intelligence safety and research company Anthropic displayed on a smartphone's screen in Brussels on June 10, 2026. Photograph: (AFP)

Story highlights

The US government has ordered Anthropic to restrict access to Fable 5 and Mythos 5 after concerns over AI jailbreaking and potential national security risks. Here's what jailbreaking means.

The US government, citing national security authorities, has directed Anthropic to suspend access to the Fable 5 and Mythos 5 by any foreign nationals, both within and outside the US territory. The direct implication of the order is that Anthorpic blocked access to the Fable 5 and Mythos 5 to all on Friday evening.

"Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5, " said Anthropic in a statement. The company now finds itself on the Pentagon's blacklist, deemed too dangerous for its own use as well as foreign use. The US Commerce Secretary, Howard Lutnick, sent a letter to Anthropic CEO Dario Amodei, saying that the model should remain locked down until the US National Security apparatus is hardened, which could take up to a few weeks.

According to the letter from the Commerce Department, Anthropic would in future require a license to export, re-export and domestic transfer of those Anthropic models. Furthermore, Anthropic will also be required to provide individually validated licenses; otherwise, it might face civil and financial penalties.

Add WION as a Preferred Source

What is "jailbreaking"?

Anthropic's newly developed frontier “AI models” have been reported to have reached a level of capability that poses a catastrophic risk to public and national security. These models can possibly find loopholes in sophisticated software codebases and generate weaponised exploits; they can perform recursive self-improvement without human oversight, and disclose information synthesising dangerous pathogens or biological weapons. So Anthropic “instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity (among others)”, jailbreaking is the possibility of bypassing those safeguards.

Trending Stories

“We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly available models are able to discover them as well without requiring a bypass,” read Anthropic's statement. It claims that no tester has been able to provide a “universal jailbreak”, something that can broadly bypass all the safeguards and unblock a wide range of cyber capabilities. It also asserts that perfect jailbreak resistance is not available. Some minor vulnerabilities are likely to remain.

Related Stories

About the Author

Share on twitter

Kushal Deb

Kushal Deb is a mid-career journalist with seven years of experience and a strong academic background. Passionate about research, storytelling, writes about economics, policy, cult...Read More

Trending Topics