Anthropic Pulls the Plug on Two Flagship AI Models
In a move that sent shockwaves through the artificial intelligence industry, Anthropic has completely cut off global access to its Fable 5 and Mythos 5 AI models — including for its own employees — after the US government issued an emergency order on Friday evening.
The order directed Anthropic to block access to the two models for all foreign nationals, both inside and outside the United States, citing national security concerns. Rather than implement a partial restriction, Anthropic opted to shut down access entirely for all customers worldwide.
What the Government Told Anthropic — and What It Didn't
In a public statement, Anthropic confirmed it was complying with the directive but pushed back on the lack of transparency surrounding it. The company said the government "did not provide specific details of its national security concern" and that any evidence of potential vulnerabilities was communicated verbally rather than through formal documentation.
Anthropig further noted that the alleged jailbreak vulnerabilities — exploits that could allow users to bypass a model's safety guardrails — were described as minor and already available through other AI models on the market. The implication: the national security justification may not be proportionate to the action taken.
The episode raises uncomfortable questions about the degree to which governments can intervene in the deployment of private AI systems, and how much notice or explanation companies are owed when such orders arrive.
Industry-Wide Implications
Fable 5 and Mythos 5 were among the most capable large language models publicly available, widely used by developers, researchers, and enterprise customers. An abrupt, total shutdown — with no advance warning — is virtually unprecedented in the AI industry at this scale.
For businesses that had integrated these models into their products, the cutoff means scrambling to find alternatives. For researchers, it means losing access to tools mid-project. And for the broader AI industry, it signals that frontier models are increasingly viewed as strategic assets — subject to export controls, national security reviews, and emergency government intervention.
Some AI policy observers have drawn comparisons to semiconductor export restrictions, where the US government has moved aggressively to limit access to advanced chips and related technology in countries deemed security risks. The Fable 5 and Mythos 5 situation suggests a similar logic may now be applied to AI model access itself.
What Comes Next
It remains unclear how long the shutdown will last or whether Anthropic will be permitted to restore access — even partially — once the government's review concludes. The company has not said whether it plans to challenge the order legally.
Anthropig's handling of the situation — complying while publicly flagging its discomfort with the opaqueness of the process — mirrors how other tech companies have navigated similar government demands in the past. It's a delicate balance: refusing to comply risks legal consequences, while silent compliance risks eroding user trust.
For now, Fable 5 and Mythos 5 are dark. And the AI world is watching closely to see what the fallout looks like — and whether other companies could be next.
Source: The Verge
