The U.S. Government has cancelled the requirement that Anthropic must obtain permission before providing Mythos and Fable models overseas. Anthropic states that access to both models will be restored from 1 July. Previously, this restriction once prevented foreign users from using the products in question.

The restrictions came into effect in June.

On 12 June, the United States Government added these two models to the list of export-restricted technologies. In accordance with that request, anthropic would have required special approval if he had provided the relevant model to foreign citizens.

This provision is difficult to implement on a large scale in practice. Anthropic thus stopped public access to the relevant models, not only to overseas markets but also to the public.

Department of Commerce says the parties have an arrangement.

Howard Lutnik, United States Secretary of Commerce, stated that Anthropic had agreed to proactively identify and address the security risks associated with the model and would work with the United States Government on the release agreements and standards for Mythos, Fable and subsequent models, while informing the Government of malicious activities.

However, in previous months, Anthropic had publicly indicated that it would voluntarily take many of these measures. As a result, some cybersecurity experts have questioned the need for this limitation, arguing that it is not necessarily primarily for security reasons.

Global competitive pressures are rising

Mythos has been open to a small number of institutions since April to reduce external concerns about their ability to identify and use software loopholes. Fable was introduced to the public in June and added additional security.

At the same time, AI Asia is also rolling out new models that are close to Mythos. It was reported that products such as Fugu and Tulonfeng were closing the gap. Against this background, the United States Government also faces pressure to ease restrictions in order to avoid being constrained in global competition by indigenous AI companies.

The industry still lacks clear rules

It was reported that recent repetitious practices of the Trump Government in relation to AI policy made it difficult for the industry to judge which rules would apply to future model publication. An executive order issued by the United States Government in June had released signals that it wished to review the model before it was released, a direction that had also been criticized by some analysts.

This means that, despite the resumption of the model visit by Anthropic, U.S. AI ' s policy expectations for product release, export and security compliance are still not fully stable.