U.S. Lifts Export Restrictions on Anthropic's Top AI Models

The U.S. government has decided to lift export restrictions on Anthropic's top-tier artificial intelligence (AI) models, Claude Mythos.

Anthropic announced on its social media on June 30 (local time) that the U.S. Department of Commerce had notified the company of its decision to lift the export control guidelines previously imposed on Mythos 5 and Fable 5.

"We will resume services starting tomorrow," Anthropic stated, adding, "We thank our users for their patience and everyone who worked to enable the redistribution of these models."

Mythos 5 is a model that shocked the global security industry by demonstrating expert-level capabilities in detecting cybersecurity vulnerabilities. Fable 5 is a version of Mythos 5 equipped with additional safety guardrails to restrict responses to sensitive topics that could be misused for hacking or weapons manufacturing.

On June 12, the U.S. Department of Commerce issued export control guidelines, citing that these two models posed a threat to national security and prohibiting foreigners from accessing them.

As Anthropic lacked the capability to identify the nationality of its users, the company blocked access to these models for everyone, including U.S. citizens.

The U.S. government reportedly issued the export control guidelines due to concerns that users could bypass the safety guardrails applied by Anthropic through methods known as jailbreaking.

The U.S. crackdown coincided with the rapid rise of Chinese open-source models, which offer performance nearly on par with top-tier U.S. models at a much lower cost. Various tech executives and investors had expressed concerns regarding the U.S. export controls, arguing that they provided Chinese developers with valuable time to narrow the technological gap.

Reported by Kim Taewon | Video by Na Hong-hee | Graphics by Lee Jeong-ju | Produced by SBS Digital News
※ Please note: This article was translated by AI and may contain errors.