Anthropic Forced to Shut Down Flagship AI Models Following U.S. Government Order

In a dramatic and unprecedented development in the global regulation of artificial intelligence (AI), leading AI laboratory Anthropic has been compelled to suspend worldwide access to its two most powerful AI models—“Fable 5” and “Mythos 5”—following an urgent directive from the U.S. government. The company confirmed on Saturday that it had complied with the restriction imposed the previous day.

Under the government order, any “foreign national,” whether located inside or outside the United States, is prohibited from accessing the two models. The restriction also applies to foreign nationals employed by Anthropic itself. As a result, thousands of developers worldwide have abruptly lost access to the services. However, older models, including Claude, will remain available.

‘Jailbreak’ Concerns Behind the Decision

While U.S. authorities cited national security concerns, they have not publicly disclosed the specific reasons behind the move. Anthropic, however, believes the decision stems from a particular “jailbreak” technique associated with the Fable 5 model—a method designed to bypass an AI system’s safety mechanisms and induce harmful behavior.

Released on June 9, the model reportedly surpassed previous benchmarks for AI intelligence. During laboratory testing, it successfully completed and defeated the video game Pokémon FireRed, a feat that attracted significant attention within the AI community. According to reports, the model’s advanced reasoning capabilities have heightened concerns among U.S. policymakers about its potential implications for cyber warfare and national security.

Anthropic Voices Frustration

In a lengthy statement, Anthropic argued that no AI model can ever be made 100 percent “jailbreak-proof.” The company said it had implemented extensive safeguards to prevent misuse of Fable 5 and criticized the government’s action as being based primarily on verbal evidence provided by an unnamed organization.

Describing the process as unilateral, Anthropic stated:

“The government should absolutely have the authority to block unsafe AI. However, such action must be part of a statutory, transparent, and clearly defined process. This sudden directive from the U.S. government did not meet those standards.”

Tightening Oversight of Silicon Valley

Technology analysts believe the timing of the government’s intervention is particularly significant. Coming shortly after OpenAI’s reported IPO filing efforts, the move against Anthropic is being viewed as a signal that Washington’s oversight of Silicon Valley’s AI industry has become more stringent than ever before.

Analysts suggest the development underscores growing concerns among policymakers regarding the rapid advancement of frontier AI systems and the potential risks they may pose to national security, cybersecurity, and global technological competition.

DBTech/BMT/OR