AI-generated abstract image of hands holding a glowing digital globe with flashing lights. — Anthropic’s Claude Fable 5 return signals a new era of global AI security collaboration and regulatory compliance.

Anthropic Restores Global Access to Claude Fable 5 Following Lifted US Export Controls

Anthropic has officially restored global access to its frontier models, Claude Fable 5 and Mythos 5, following the lifting of U.S. government export controls on June 30, 2026. The redeployment follows a period of intense collaboration with the U.S. government to implement advanced safety classifiers and a new industry-wide framework for assessing AI “jailbreak” risks.

RMN Digital Corporate Desk
New Delhi | July 5, 2026

San Francisco, CA — Anthropic has announced that access to its most advanced AI models, Claude Fable 5 and Claude Mythos 5, has been restored after a brief but significant disruption caused by federal export controls. Starting Wednesday, July 1, Fable 5 will be available globally across the Claude Platform, Claude.ai, and integrated tools like Claude Code.

The Regulatory Hurdles

The disruption began on June 12, when the U.S. government applied immediate export controls to these frontier models. Because Anthropic lacked a real-time method to verify the nationality of its users, it was forced to suspend access for all users globally to remain compliant with restrictions regarding foreign nationals. These controls were officially lifted on June 30, clearing the path for the current redeployment.

Defense in Depth: Anthropic’s new safety classifiers block 99% of reported bypass techniques, ensuring frontier power stays within safe boundaries.

Addressing Safety Concerns

The government’s initial directive was prompted by a report from Amazon researchers who discovered a method to bypass Fable 5’s safeguards. The bypass allowed the model to identify software vulnerabilities and, in one instance, demonstrate how an exploit could be produced.

In response, Anthropic worked closely with the Center for AI Standards and Innovation (CAISI) to develop an improved safety classifier. This new mechanism is designed to block the specific techniques described in the Amazon report in over 99% of cases. While the new safeguards are “extraordinarily strong,” the company noted they may increase false positives during routine coding and debugging tasks as part of a “safety margin” approach.

A Shared Bar: Leading AI developers unite to standardize jailbreak severity assessments, bringing much-needed clarity to AI risk management.

A New Industry Standard

Beyond technical updates, Anthropic is spearheading a consensus framework to standardize how the industry assesses AI “jailbreaks”—techniques used to bypass model safeguards. Partnering with tech giants including Amazon, Microsoft, and Google, the company is proposing a four-point scoring system based on:

Capability gain
Breadth of capability
Ease of weaponization
Discoverability

Deepening Government Ties

Anthropic has committed to a deeper level of transparency with the U.S. government. This includes providing designated partners with early access to models for independent evaluation before broad release, rapid information sharing regarding significant jailbreaks, and dedicated resources for joint research on AI security.

“Our hope is that this collaboration… will serve as the basis for systematic rules for the whole industry,” the company stated, emphasizing the need for durable, transparent processes in AI releases.