Anthropic blocks all Mythos access after US order to cut off non-citizens

Anthropic said Friday that it suspended all access to its most cyber-capable AI models, Mythos 5 and Fable 5, citing a US government order it received to oust all non-US citizens from access.

According to Anthropic, the US government has been made aware of ways to bypass security safeguards in Fable 5, likely prompting Washington’s national security concerns.

The company says it has “abruptly cut” all access, including to US citizens, as a result of the order, though it disagrees with the government decision.

Back in April, Anthropic released Mythos to a select few organisations, none in the EU, over claims that the model showed unprecedented capabilities to detect and exploit cybersecurity flaws in systems.

But the limited launch raised alarm bells among governments and critical industries, including in Europe, that did not get access, and feared this would leave them unprepared for a wave of cyberattacks when such models would eventually come on the market.

Fable 5 was launched later and is a widely accessible version of Anthropic’s cyber-capable model, but it includes safety backstops.

In the EU, the US company’s decision to limit Mythos’ launch had reignited calls for European sovereignty. Back then, lawmakers from the European Parliament wrote to the Commission, asking what its plans were to prepare for the impending “cybergeddon” and to develop Europe’s own sovereign AI cyber-capabilities.

The EU’s cybersecurity agency ENISA was offered access in early June, following weeks of talks between the Commission and Anthropic.

Most advanced AI models are set to pose a major risk to cybersecurity, and other AI model providers from the US and China are expected to not be far behind Anthropic. OpenAI has already launched its own cyber model soon after Mythos came out.

Anthropic said in its Friday statement that it believes the US government’s concerns over ways to bypass Fable 5 aren’t warranted, as the model’s capabilities even then remain in line with others on the market.

“We believe this is a misunderstanding and are working to restore access as soon as possible,” Anthropic’s statement reads.

(at)