The US lifts restrictions on Anthropic’s two advanced models
Anthropic stated it will soon restore global access to its Fable 5 and Mythos 5 models after the US government removed the barriers.
“We received notification that the US Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We will begin restoring access starting tomorrow,” Anthropic announced in a post on X on June 30.
The announcement came four days after the AI company was permitted by the US government to open access to the Mythos 5 model to a small group of cybersecurity companies in the country.
In a letter dated June 26 and collected by Reuters, US Commerce Secretary Howard Lutnick stated that Anthropic had cooperated with the US government to address “threats related to the models mentioned.”
Claude Fable 5 Logo. Photo: Anthropic
OpenAI last week also complied with government requests and restricted access to its GPT-5.6 model, only allowing some approved partners to use it. “This is not a process we consider optimal,” CEO Sam Altman said.
Director of the US Central Intelligence Agency (CIA) John Ratcliffe compared the capabilities of advanced AI models to nuclear weapons, while also defending the stringent control measures of President Donald Trump’s administration regarding the rollout of the country’s most powerful AI technologies.
Anthropic announced Fable 5 and Mythos 5 in early June, two models advertised by the company as the most modern across various metrics in the AI industry. They were built upon the Claude Mythos Preview platform, which previously concerned US officials due to its advanced cybersecurity capabilities.
By June 12, Anthropic received directives from the US government to suspend access to Fable 5 and Mythos 5 for “foreign nationals, whether in the US or another country, including Anthropic employees.” The AI startup stated that the US government cited “national security” as the reason, but did not provide details.
David Sacks, co-chairman of the US President’s Council of Advisors on Science and Technology, stated that “a trusted partner” of the US government and Anthropic tested Fable and discovered a jailbreaking method to bypass the safety barrier that was intended to separate this commercial model from Mythos’s unrestricted cybersecurity capabilities.
The US government then requested Anthropic to patch the vulnerability or recall Fable 5 until the issue was resolved, but CEO Dario Amodei reportedly refused. “The administration then reluctantly issued export control directives. They are ready to lift the restrictions as soon as the problem is fixed,” he said.
Sacks criticized Anthropic’s actions as inconsistent with its position as a laboratory that always prioritizes safety, and also for having previously called for Mythos to be managed as a cybersecurity weapon.
He stated that the restriction decision was not related to previous friction between Anthropic and the government, adding that Washington always values the AI startup’s technology and considers the issue “very easily resolvable.”
Diep Anh (According to AFP, CNBC)
