Insiders say that Amazon's security team used Anthropic's newly released Fable 5 AI model to find (information security vulnerabilities that should not have been provided), just like a sister finding out her brother is eating snacks secretly and quickly running to daddy to tattle, so the naughty son is punished and the users suffer.

Third-party individuals who have seen relevant reports say that the vulnerabilities discovered by Amazon are related to Defense - Oriented Prompt Interception (DOP), and are far from jailbreaking.

Amazon's researchers, by using a series of prompts, made Fable 5 reveal to users vulnerability information that could be used for cyberattacks. But Anthropic believes that these vulnerabilities have long been known to the company and their harm is negligible. Third - party individuals think that these so - called vulnerability information is actually more useful to defenders, and it is completely overreacting on the part of the White House.

After receiving the report from researchers, Amazon CEO Andy Jassy called government officials to express his concerns, but the matter quickly evolved into the Commerce Department's comprehensive ban on foreign users using Mythos and Fable. Critics believe that the normal procedure would be for Jassy to quietly settle the matter with Anthropic instead of stirring things up with the White House. Is Amazon trying to pressure Anthropic, warning it to be loyal to AWS instead of seeking computing power from Google and Musk?

The commentariat believes that the US government has a deep - seated prejudice against Anthropic. National Cybersecurity Director Shawn Kernick and Commerce Secretary Howard Kurtz are both involved. Without providing written evidence, just based on the tattle - tale of a "little cotton - padded jacket", they have shackled Anthropic with export control.

Since some of Anthropic's employees are foreigners, it is not realistic to comply with this directive, so they adopt a "one - size - fits - all" approach and no one can use it. In the company's public letter explaining the comprehensive shutdown, it is stated that for this level of vulnerability, there is no need to deploy Mythos and Fable, nor is there any need to compete with Fable 5 in a roundabout way. Competitors' models (such as OpenAI's GPT - 5.5) can directly provide similar information.

Source: Quick Technology