Anthropic Resumes Global Service for Claude Fable 5 Amid Eased AI Export Restrictions

Claude Fable 5 Resumed Service, Captured from Anthropic's Official X Account. [Photo=Yonhap News]

Anthropic, the American artificial intelligence (AI) company, has resumed global services for its high-performance AI model, Claude Fable 5. This decision follows the U.S. government's easing of certain export restrictions imposed due to security concerns. However, the model known as Mythos 5, which is classified as more sensitive due to its vulnerability detection capabilities, is expected to remain limited to select U.S. institutions for the time being.

On July 1, Anthropic announced the resumption of the Fable 5 service via its X account (formerly Twitter). As a result, users can once again access Fable 5 through Anthropic services, including Claude and Claude Code.

Fable 5 is Anthropic's high-performance public model, comparable in capabilities to Mythos 5, which excels in analyzing cybersecurity vulnerabilities. It is designed to block requests that could be misused for hacking or the production of biological and chemical weapons.

Previously, the U.S. government raised concerns that both models could be exploited for cyberattacks, particularly noting that bypassing Fable 5's safety features could lead to the misuse of its software vulnerability detection capabilities.

In response, Anthropic enhanced its model to filter out risky requests, stating, "We have improved the model to block over 99% of reported risky behaviors."

While Fable 5 is now available to users worldwide, Mythos 5 remains under separate control. Bloomberg reported that the U.S. government is restricting access to Mythos 5 primarily to certain domestic institutions.

As a result, the expansion of Anthropic's security vulnerability detection collaboration program, Project Glasswing, is likely to be delayed. In South Korea, companies such as Samsung Electronics, SK Hynix, and SK Telecom have been mentioned as participants in this program, but the timeline for utilizing Mythos 5 depends on further approvals from the U.S. government.

This situation highlights the U.S. government's increasing direct management of high-performance AI models from a national security perspective. As AI models become subject to export controls similar to semiconductors and advanced equipment, the ability of foreign companies to utilize advanced AI is increasingly influenced by U.S. approval processes.

* This article has been translated by AI.

AJP