Anthropic AI Models Suspended Over Cybersecurity Fears Reported by Amazon

The boss of Amazon has reported a serious security problem to the United States government. Andy Jassy, the chief executive of Amazon, spoke with Treasury Secretary Scott Bessent. He warned the government about two artificial intelligence models from Anthropic. These models are called Fable 5 and Mythos 5.

Researchers at Amazon tested the Fable 5 model recently. They found a major weakness in the system. The AI model provided dangerous information that hackers could use for cyberattacks. Because of this discovery, the government has created a new rule. Officials banned the export of both Fable 5 and Mythos 5 to other countries. As a result, Anthropic stopped worldwide access to these models on Friday.

An Amazon spokesperson has confirmed part of the story. The spokesperson explained that governments often ask Amazon for advice about security risks. However, the company refused to share the specific details of their private conversations. The block on the Anthropic models has also caused problems for Amazon's own cloud computing services.

David Sacks, a technology advisor for the government, shared his view on the situation. He said that a trusted partner discovered a way to break the AI's safety rules. The government asked Anthropic's chief executive, Dario Amodei, to fix the problem. According to David Sacks, Amodei decided to reject the request.

Anthropic has published a response on their official blog. The company strongly disagrees with the government's decision. They argue that the dangerous capabilities already exist in other AI models. If people want to find this information, they will just use a different program. The debate about AI safety continues to grow.