Analysis: Anthropic refuses to patch Fable jailbreak vulnerability, leading to US government imposing export controls

BlockBeats News, June 14th: The White House AI and Crypto Czar, David Sacks, posted a message stating that the commercial version of the Mythos series model, Fable, released by Anthropic this week, has added a security fence. However, once the fence is bypassed, users will be able to access Mythos' advanced cyber attack capabilities. Sacks pointed out that Anthropic has always described Mythos as a "supervised network weapon," so fixing the related vulnerabilities should have been its responsibility.

Sacks stated that a partner trusted by both Anthropic and the U.S. government discovered a jailbreak method that could bypass the security fence while testing Fable. The U.S. government then requested Anthropic CEO Dario Amodei to fix the vulnerability or take the model offline, but the request was denied. In a statement, Anthropic considered the vulnerability "not serious," a claim that contradicted the assessments of the U.S. government and relevant partners.

Sacks expressed that Anthropic has always emphasized that security should be a top priority but appeared to prioritize maintaining consumer-grade model services this time. In response, the U.S. government reluctantly imposed export control measures on Anthropic and hoped that Anthropic would quickly address the security issues to lift the restrictions and resume the full release of Fable.

Sacks also denied that this action was related to the previous dispute between the U.S. Department of Defense and Anthropic, stating that the government recognizes Anthropic's technical capabilities and believes that the current issue can be easily resolved, with Anthropic currently holding the initiative.