The White House administration has decided to impose export restrictions on AI models from Anthropic — Fable 5 and Mythos 5. The key trigger was concerns that a group linked to China may have gained access to the Mythos model. An additional factor was reports of a possible bypass of Fable 5's security mechanisms. Anthropic itself insists that it was a "narrow" and non-universal method that did not give the model unique capabilities.
What exactly happened?
According to my data, the Donald Trump administration was alarmed by suspicions that an organization with ties to the PRC had accessed Mythos. Details — which group, how it happened, and where the White House obtained the information — remain undisclosed. Notably, an Anthropic representative in talks with authorities denied that the topic of "Chinese access" to Mythos was ever raised. The company emphasizes that access to its products from China is prohibited.
However, another line of events was unfolding in parallel. On June 11, Amazon CEO Andy Jassy informed the administration about a possible way to bypass Fable 5's restrictions. This occurred during a model review. As a result, U.S. Secretary of Commerce Howard Lutnick sent a letter to Anthropic CEO Dario Amodei, ordering Fable 5 and Mythos 5 to be subject to export controls. Officials tried to persuade the company to voluntarily suspend access to the models but were refused. According to a version close to the startup, the White House gave just 90 minutes to shut down access, without providing details about the threat.
White House advisor David Sacks later stated that the administration received a warning from a partner trusted by both sides. He expressed hope that Anthropic would fix the security issue, after which the export controls would be lifted and Fable would return to public access. "The ball is in Anthropic's court," Sacks concluded.
Anthropic's position and my conclusions
Anthropic strongly disagrees with this interpretation. The company claims that the authorities' letter contained no specific details about risks to national security. The startup studied the bypass technique and concluded that it only allows finding "a small number of previously known minor vulnerabilities." Moreover, other publicly available models can identify the same issues without bypassing security. Before launching Fable 5, Anthropic spent "thousands of hours" testing safeguards jointly with the U.S. government, the UK AI Safety Institute, and third-party organizations. Testers found no universal way to massively bypass the model's restrictions. The company considers the incident a misunderstanding and is working to restore access.
Recall that Fable 5 and Mythos 5 were unveiled on June 9. Fable 5 is a Mythos-class model for public access, while Mythos 5, with relaxed restrictions, is intended for cybersecurity defenders under Project Glasswing. On June 2, Anthropic expanded this program to approximately 150 organizations from 15 countries, and early partners have already found over 10,000 high and critical severity vulnerabilities.
My analysis: This incident is a vivid example of how geopolitical risks are beginning to directly impact the AI market. The White House's reaction appears excessive and hasty, especially given the lack of concrete evidence. Anthropic, with its emphasis on security, clearly did not expect such pressure. In the short term, this creates chaos for developers using Fable 5 and undermines trust in U.S. AI products. However, for decentralized AI projects like Venice and Morpheus, whose tokens rose by 14% and 21% following this news, the situation conversely opens a window of opportunity. The market signals that centralized solutions are becoming vulnerable to political interference.