In brief: Following export controls due to security concerns in vulnerability identification, Claude Fable 5 will be released globally starting July 1, 2025, with newly assessed safeguards.
The US government lifted export controls on Anthropic’s Claude Fable 5 and Claude Mythos 5 on June 30, which had been in place since June 12. Starting July 1, Fable 5 is available globally again after a vulnerability analysis required security adjustments.
On June 12, US authorities imposed export controls on Claude Fable 5 and Claude Mythos 5 after Amazon researchers discovered a method to bypass Fable 5’s security mechanisms. The technique enabled the model to identify software vulnerabilities and in one case produce code demonstrations of their exploitation. Anthropic had to immediately block access for all users, as reliable real-time verification of nationality was not possible.
Tests conducted by Anthropic and the US government showed that less capable models — including Claude Opus 4.8, GPT-5.5, and Kimi K2.7 — could identify the same vulnerabilities as Fable 5. When demonstrating the exploitation of a single vulnerability, all tested models (Claude Haiku 4.5, Sonnet 4.6, various Opus versions up to 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7) produced the same results as Fable 5. The analysis confirmed that the technique did not reveal unique Mythos-level cyber capabilities, but rather represented an edge case in Fable 5’s safeguards.
Starting July 1, Fable 5 is available globally on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. For Pro, Max, Team, and select Enterprise plans, Fable 5 is included free of charge for up to 50 percent of weekly usage limits until July 7, after which it is available via Usage Credits. Access via AWS, Google Cloud, and Microsoft Foundry will be restored as quickly as possible. Mythos 5 was released for US organizations following government approval on June 26; coordination to expand to additional Glasswing partners is underway.
Anthropic is developing a unified assessment framework for security vulnerabilities and so-called jailbreaks in AI models together with Amazon, Microsoft, Google, and other Glasswing partners. The goal is a consistent standard for severity assessment that helps AI developers prioritize findings, deploy powerful models more securely, and communicate risks uniformly to authorities and partners. Anthropic is also strengthening collaboration with the US government through pre-release testing, information sharing, and research cooperation.
Source: www.anthropic.com · Published June 29, 2026
Lumi AI News — AI-assisted curation in accordance with Article 50 EU AI Act. Paraphrasing and classification via Lumi News Pipeline v1.7.2.