Anthropic has announced the global redeployment of its artificial intelligence model, Claude Fable 5, starting Wednesday, July 1, 2026. This decision follows the lifting of US government export controls that had temporarily suspended access to the model on June 12, 2026.
The export controls also affected Claude Mythos 5, a model sharing the same underlying architecture but configured with fewer safeguards for defensive cybersecurity tasks. While Mythos 5 access was restored to select US organizations on June 26, Claude Fable 5 is now returning to global users across various platforms.
The Export Control Timeline
The temporary suspension began after the US government issued an immediate directive on June 12. This order was prompted by an Amazon research report showing that the model could identify software vulnerabilities and demonstrate an exploit.
Because Anthropic could not verify user nationalities in real-time to comply with the export restrictions, the company suspended access for all users. Over the subsequent two weeks, the company worked with government officials to implement updated safety classifiers.
Updated Safeguards for Claude Fable 5
To address the vulnerability exploit issue, Anthropic trained an improved safety classifier that blocks the specific bypass technique in over 99% of cases. If a user request to Claude Fable 5 is blocked by this classifier, the system will automatically redirect the query to the Claude Opus 4.8 model.
The company noted that these safety margins may occasionally flag benign requests during routine coding and debugging. However, this defensive approach is designed to prevent the model from being used for offensive cyber operations.
A Shared Industry Framework
In response to these events, Anthropic is collaborating with Amazon, Microsoft, and Google to develop a standardized framework for assessing AI jailbreaks. This proposed system will evaluate jailbreaks based on capability gain, breadth, ease of weaponization, and discoverability.
Furthermore, the company is launching a new HackerOne program to allow security researchers to submit potential vulnerabilities. This initiative aims to identify and mitigate risks before they can be exploited by malicious actors.
Access and Future Outlook
Starting July 1, users can access the model on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork apps & software. Pro, Max, Team, and select Enterprise plans will include the model for up to 50% of weekly usage limits through July 7, after which usage credits will be required.
Anthropic is also working to restore access on AWS, Google Cloud, and Microsoft Foundry. The company plans to deepen its collaboration with the US government through pre-release evaluations and rapid information sharing.
Source: anthropic.com
