The deadlock is broken, and the United States lifts export controls on Anthropic's two flagship AI models.

The US government has officially revoked the export controls on two top - tier models under Anthropic, ending a weeks - long standoff between the Trump administration and the AI enterprise valued at trillions of dollars.

Industry insiders revealed that the US Department of Commerce notified Anthropic on the evening of June 30 (Eastern Time) to revoke the global export controls on Anthropic's two flagship large models, Claude Fable 5 and Claude Mythos 5. Anthropic officially stated that the two models will resume global access in batches starting from July 1.

One week prior, the regulatory authorities had relaxed some restrictions in advance, allowing Anthropic to re - release the Mythos 5 model to about 100 partner enterprises that had undergone strict qualification reviews. This version of the model has fewer security protection mechanisms configured and is mainly targeted at enterprise - level commercial customers.

Some media pointed out that in an official letter issued to Tom Brown, the co - founder of Anthropic, Commerce Secretary Lutnick stated that the core prerequisite for lifting the controls is that Anthropic promises to actively identify and address various security risks related to its models.

Origin of the Controls

This export ban started on June 12 and was signed into effect by Lutnick. Previously, US regulatory agencies found that the outside world had mastered a "jailbreaking" method that could bypass the built - in security protection mechanisms of the models, posing serious security risks. After the controls were implemented, Anthropic was forced to recall all high - performance large - model products globally.

The above - mentioned ban only targeted the two latest flagships, Fable 5 and Mythos 5. Mid - and low - end models such as Claude Opus 4.8 and the Sonnet series were not restricted at all throughout the process and remained open for global access as normal.

This incident significantly intensified the contradiction between the White House and the Silicon Valley technology industry. The Silicon Valley industry criticized the US government for its ad - hoc and non - unified standard control model for cutting - edge AI. At the same time, governments of many countries publicly expressed their dissatisfaction, accusing the US of unilaterally issuing the ban, which indirectly damaged the interests of its allies.

The background for the introduction of this export control is rather contradictory: under the pressure of the AI industry lobbying group, the Trump administration previously introduced a voluntary regulatory framework for cutting - edge large models and publicly promised not to establish a mandatory and regular regulatory system for leading AI enterprises.

However, many high - level US government officials have always maintained a warning attitude: if high - performance large models are opened to the whole society without fully assessing the risk of the models being used to exploit network security vulnerabilities in key industries, it will bring unpredictable security threats.

The First in the Industry

Anthropic has achieved business overtaking with its self - developed series of models, surpassing competitors such as OpenAI in the industry. It is also the first AI enterprise to be subject to an export ban from the White House due to the security risks of its models being misused by malicious forces.

OpenAI, also a leading AI manufacturer, is also subject to regulatory constraints for its new model GPT - 5.6. The Trump administration requires the company to initially only grant the right to use the model to more than twenty partner enterprises approved by the official. According to insiders, GPT - 5.6 will be opened to a wider range as early as next week.

Industry insiders reported that while the US is relaxing the model controls on leading enterprises, the US and Europe have started consultations and plan to establish a "trusted partners" program for cutting - edge AI models. According to this program, the core allies of the US will have priority access to the latest AI models.

Anthropic officially stated that it would not comment on the lifting of this export control.

This article is from the WeChat official account "Caixin AI Daily". Authors: Li Ying and Qi Ling. Republished by 36Kr with authorization.