Anthropic had 90 minutes to restrict Claude Fable 5 as White House feared Chinese access

On June 12, a swift directive from the White House compelled Anthropic to impose restrictions on its advanced AI models, specifically the Claude Fable 5 and Mythos 5. The government mandated that these models only be accessible to U.S. citizens, significantly curtailing global availability. Reports indicate that Anthropic was given a mere 90 minutes to comply with this export-control order. The urgency behind this decision stemmed from concerns that a group potentially linked to China might have gained unauthorized access to the Mythos AI model. While the specific details of how this access was achieved or which entity was involved remain undisclosed, the implications are serious. Access by the Chinese government to such sophisticated technology poses a potential risk to U.S. national security, particularly with fears surrounding the possibility of reverse-engineering the model. Following the directive, Anthropic expressed frustration, stating that the White House had not provided a comprehensive explanation in advance. This lack of clarity compelled the company to send technical executives to engage with government officials, aiming to clarify the situation and address any concerns. Earlier in April, Anthropic had introduced the Mythos Preview under stringent access restrictions, allowing only a select number of companies and government entities to utilize the model due to its potential for misuse. Additionally, David Sacks, a former adviser to President Trump, provided insights via a social media post regarding the events leading up to the White House's decision. He noted that officials were warned about a possible vulnerability in the Fable 5 model that could allow it to be 'jailbroken,' meaning individuals could circumvent its safety protocols. In conversations with various stakeholders, Sacks concluded that Anthropic's leadership, including CEO Dario Amodei, downplayed the severity of this flaw, asserting it was not a significant threat. However, he claimed that the company had declined to address the issue. This situation is compounded by similar jailbreak vulnerabilities reported by Amazon, which led to discussions between Amazon's CEO, Andy Jassy, and U.S. officials. As a result of these growing concerns, Anthropic has now restricted access to all users, not just those outside the U.S. The company is scheduled to meet with White House representatives in Washington D.C. on June 22, 2026, to discuss these matters further.