US Eases Restrictions, Anthropic’s Advanced AI Models Set to Be Reopened

Staff Report: PNN

After the easing of US government export restrictions, Anthropic’s advanced artificial intelligence (AI) models Mythos and Fable are once again being made available to international users. The company announced that access to these models will be gradually restored starting Wednesday (July 1).

On June 12, the US government included Mythos and Fable in its export-controlled technology list. As a result, these models could not be provided to foreign users without special government approval. In practice, the complexity of obtaining such approvals made it difficult for the company to continue normal access, forcing Anthropic to temporarily suspend availability for general users.

However, after several weeks of discussions, the situation changed. US Commerce Secretary Howard Lutnick stated that Anthropic will actively identify and address safety risks related to its models in the future. The company will also cooperate with the government to establish safety standards and protocols before releasing new models, and report any misuse promptly to authorities.

Anthropic stated that most of these commitments were already voluntarily followed by the company. Many cybersecurity experts had questioned the justification of the restrictions from the beginning, arguing that political and policy pressures played a larger role than actual security concerns.

In April this year, Mythos was initially released in a limited form for selected organizations. Due to concerns over its ability to identify and analyze software vulnerabilities, its use was initially restricted. Later in June, after additional security measures were added, the Fable version was released for general users.

Meanwhile, several Asian technology companies have rapidly begun launching advanced AI models. New models such as Fugu and Tulongfeng are approaching Mythos in technical capability, increasing pressure on the United States to maintain its competitive position in the global AI race.

According to analysts, this competitive landscape also played a key role in the decision to ease restrictions. Earlier last week, Mythos access was granted to selected approved customers under White House authorization. A similar policy was also applied to OpenAI’s latest AI models, allowing limited access to approved institutions.

However, uncertainty over US AI policy remains. A recent executive order indicated possible government review before releasing new AI models, which has drawn criticism from many technology experts and policymakers.

Experts believe that balancing AI safety, export control policies, and global competition will remain one of the biggest challenges for the US government in the coming years.