Another major tech player appears to be behind the sudden worldwide ban of Anthropic's new AI model Fable: Amazon CEO Andrew Jassy reportedly informed US Treasury Secretary Scott Bessent, among others, about a suspected jailbreak, leading the US government to issue an export directive that in turn forced Anthropic to ban the AI model.
Several media outlets are reporting consistently on the case. According to these reports, Andrew Jassy first had his cybersecurity department test the new AI model for possible jailbreaks – tricks used to make an AI model process prompts that it would otherwise refuse for security reasons. The experts reportedly presented a report indicating that the AI model provided working exploits of vulnerabilities in four common programs. Jassy then also informed US government officials about this. An Amazon spokesperson confirmed to The Wall Street Journal that the US government often seeks advice from Amazon as a leading tech company on technology matters, but declined to comment on specific individual cases.
Sacks: US government acted “reluctantly”
In response to the report, the US government gave Anthropic an ultimatum: the jailbreak issues had to be resolved within 24 hours – or access to Fable would be blocked for non-Americans. David Sacks, a tech investor himself and co-chair of the US President's Science and Technology Advisory Council, reported on X how this allegedly happened. According to his account, Anthropic CEO Dario Amodei refused to fix the security vulnerabilities identified by Amazon. Previously, Anthropic had denied in a statement that it was a full jailbreak, stating that the tasks performed are also easily achievable with OpenAI's competitor model ChatGPT 5.5. Sacks' account claims the US government “reluctantly” imposed the strict ban due to Amodei's refusal. The hope is still that Anthropic will fix the alleged problems and the model can be used normally again. Meanwhile, Anthropic can only comply with the requirements by blocking its new AI model worldwide.
