Industry media AXIOS reported that Anthropic has sent a senior technical team to Washington to conduct emergency consultations with U.S. government officials, with the goal of lifting the export control ban on the Claude Fable 5 model as soon as possible. The ban, issued by the U.S. government last week, requires Anthropic to immediately cut off access to Claude Fable models to all foreign citizens, even those in the United States.
Due to compliance considerations, Anthropic directly removed Claude Fable from the shelves globally. At the same time, Anthropic also expressed strong dissatisfaction with the export control ban issued by the US government. However, the problem has to be solved, so Anthropic can only send a team to communicate with the US government in the hope that the ban can be lifted as soon as possible and normal access to the model can be restored.
However, there is currently no timetable. If the negotiations go well, the U.S. government may lift the export control ban. Anthropic may solve the security issues that the U.S. government is concerned about by adding a layer of stricter security restrictions. The security issue here refers to jailbreaking the model, thereby bypassing the original network security boundary and allowing the model to give more answers about sensitive content.
It is reported that the Amazon security team was the one who reported the network security issues of Claude Fable to the U.S. government. This was the direct reason for the U.S. government to immediately issue an export control ban. Previously, AI jailbreak experts also publicly posted about the Claude Fable jailbreak method on social media. It is not known whether Amazon’s security team implemented the model’s jailbreak with reference to the solution shared by the AI jailbreak experts.