Anthropic cuts access to AI models after US order
Anthropic cuts access to AI models after US order
https://www.taipeitimes.com/News/biz/archives/2026/06/15/2003859091
Publish Date: 2026-06-14 12:00:00
Source Domain: www.taipeitimes.com
Here is an unordered list of key points from the article:
* Anthropic PBC temporarily disabled access to two AI models, Fable 5 and Mythos 5, in response to a government directive aimed at preventing the models from falling into the hands of foreign nationals for national security reasons.
* The directive was issued following concerns that the Fable 5 model could be “jailbroken” to aid hacking, though Anthropic believes this method doesn’t provide capabilities unavailable elsewhere.
* Fable 5 is a locked-down version of Mythos 5, which is a cutting-edge AI model that comes with significant national security implications due to its ability to identify software vulnerabilities.
* The restricted model Mythos 5 has only been released to select companies, and the EU recently secured access after lengthy negotiations.
* The European Commission commented that this development highlights the importance of “technological sovereignty” for Europe.
* Researchers at Amazon.com Inc. had previously discovered vulnerabilities in Anthropic’s model, and Amazon had communicated these findings to the US government prior to the directive.
* Anthropic maintains that there is no “universal jailbreak” and disagrees with the decision to disable the models due to the perceived risks.
This summary outlines the primary points discussed in the article, while maintaining respect and adherence to the Media Outlets Policy.
