Anthropic has restored global access to its flagship Claude Fable 5 AI model after the US Department of Commerce lifted emergency export controls that had forced the company to suspend the platform for nearly three weeks. The model officially returned on July 1, making it available again through Claude.ai, the Claude Platform, Claude Code and Claude Cowork, while access on Amazon Web Services (AWS), Google Cloud and Microsoft Foundry is also being restored. The move ends one of the first major cases in which a commercial AI model was temporarily taken offline because of national security concerns.
The issue began just 3 days after Fable 5 and Mythos 5 were launched on June 9. On June 12, the US government imposed export controls after learning of a report from Amazon researchers describing a way to bypass, or ‘jailbreak’, some of Fable 5’s cybersecurity safeguards. The technique allowed the model to identify certain software vulnerabilities and, in one case, generate proof-of-concept exploit code showing how a vulnerability could be used. The government ordered Anthropic to block access for all foreign nationals. However, because the company could not reliably verify every user’s nationality in real time, it decided to suspend both Fable 5 and Mythos 5 for all users worldwide.
During the shutdown, Anthropic worked with the US Department of Commerce and the Center for AI Standards and Innovation (CAISI) to improve the model’s safety. Instead of changing the core AI model, the company introduced a new AI-powered safety classifier that specifically detects prompts linked to the reported jailbreak. If such a prompt is identified, the request is blocked and automatically redirected to the older Claude Opus 4.8 model. Anthropic claims that the updated system blocks the reported jailbreak in more than 99% of test cases.
Anthropic has argued that the reported issue was not unique to Fable 5. According to the company, its internal testing found that several other leading AI models, including GPT-5.5 and Kimi K2.7, could identify the same software vulnerabilities and produce similar demonstration exploits. It also said the reported jailbreak did not unlock the more powerful ‘Mythos-level’ cybersecurity capabilities. Instead, it exposed a narrow case involving routine defensive cybersecurity tasks. Anthropic maintained that if every isolated jailbreak led to a model being withdrawn, it would become extremely difficult for any company to release frontier AI models because no advanced system is completely resistant to carefully designed prompt attacks.
Although Fable 5 has returned for general users, Mythos 5 remains under much tighter controls. The model is currently available only to a limited number of vetted US organizations involved in cybersecurity and critical infrastructure through Anthropic’s Project Glasswing programme. Along with the relaunch, the company announced several new safety measures, including a HackerOne bug bounty programme to encourage researchers to report jailbreak techniques responsibly, more extensive pre-release safety testing, and closer cooperation with the US government.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →
Ashutosh is a Senior Writer at The Tech Portal, largely reporting on new tech, and intersection of technology and business. Ashutosh’s career spans across nearly a decade of technology writing across multiple platforms and languages.
