Anthropic reopens global access to Claude Fable 5its most advanced artificial intelligence model, after the United States Government will lift export controls that he had recently imposed on this system and on Claude Mythos 5.
The company has announced that Claude Fable 5 is available again for users around the world starting this Wednesday, July 1. The decision comes after several weeks of restrictions that forced Anthropic to block access to both models for foreign citizens, both inside and outside the United States, including the company’s own workers. Since the government order took effect immediately and Anthropic had no way to implement it, the company chose to suspend access to Fable 5 and Mythos 5 for all its users.
The ultimate reason is Mythos 5’s ability to find software vulnerabilitieswhich makes it dangerous in the wrong hands and hence it has had limited access from the beginning, only from approved companies and organizations. Anthropic released Fable 5 as a version of Mythos with safeguards allowing it to be used by the general publicbut it was soon shown that they were not enough, leading the Trump Administration to make that decision.
The US Secretary of Commerce, Howard Lutnickassured this Tuesday that The Government had worked for the last two weeks with Anthropic to analyze and approve Fable 5with the aim of guaranteeing coordination within the Administration and reinforcing the United States’ leadership in artificial intelligence.
With restrictions now lifted, Claude Fable 5 is available again on Claude Platform, Claude.ai, Claude Code and Claude Cowork. Anthropic has explained that users Pro, Max, Team and some clients Enterprise will have access to Fable 5 until 50% of your weekly usage limits until July 7. From then on, the AI model will work through usage credits.
The company also plans to revive Fable 5 in AWS, Google Cloud and Microsoft Foundry as soon as possible.
The situation of Claude Mythos 5 is different. Anthropic says it has restored access to the model for a group of US organizations after receiving government approval on June 26. However, the company continues negotiating with the authorities to expand access to more national and international partners within its program. Glasswing.
Anthropic has also given more details about the origin of the restrictions. According to the company, Amazon researchers found a way to bypass Fable 5’s safeguards to make the model identify software vulnerabilities. In one case, the system generated code that showed how one of these vulnerabilities could be exploited.
The company clarifies, however, that other models, including Claude Opus 4.8, GPT-5.5 and Kimi K2.7, could also identify the same vulnerabilities.
To correct the problem, Anthropic has trained a new security classifier specifically designed to block the behavior described in the Amazon report. According to the company, the specific route used to bypass the protections is now blocked in more than 99% of cases.
When a request to Fable 5 is rejected by this new security system, the user will receive a notice and the request will be redirected to Claude Opus 4.8.
Anthropic also works with Amazon, Microsoft, Google and other partners of the Glasswing program into a common framework for Assess the severity of methods used to bypass security protections for AI models. This framework will take into account factors such as the increase in capabilities that the attack allows, the extent of that increase, the ease of converting it into an offensive tool, and how easy it is to discover it.