In line with the Wall Road Journal, the export management directive that led to Anthropic chopping off entry to Fable 5 and Mythos 5 was triggered partially by cybersecurity analysis from Amazon and conversations between CEO Andy Jassy and the White Home. In line with the report, the paper from Amazon claims that, via a sequence of prompts, it was capable of get Fable 5 to serve up info that might be utilized in cyberattacks. Amazon has but to answer a request for remark.
Shortly after Jassy shared the corporate’s findings with the federal government, it made the decision to dam its use by overseas nationals. Complicating this difficulty is that lots of Anthropic’s researchers are foreign-born, that means they had been barred from accessing their very own product.
In a press release, Anthropic disputed the federal government’s characterization of the difficulty as a “jailbreak.” It argued that most of the similar vulnerabilities might be found utilizing different publicly accessible fashions, together with GPT 5.5. Some safety researchers seem to again the corporate’s interpretation. Katie Moussouris, the founder and CEO of LutaSecurity posted on BlueSky that “I’ve seen the paper. It’s not a jailbreak.” Former Commerce Division official Kate Koren purported to the WSJ that the White Home’s dislike of Anthropic might have influenced the choice.
Anthropic and the Trump administration have been at odds for a while over the corporate’s refusal to permit its AI for use for mass surveillance of People or to energy deadly autonomous weapons. In February, Trump instructed federal businesses to cease utilizing Anthropic’s AI. And simply hours later, Secretary of Protection Pete Hegseth designated the corporate a provide chain danger.
The federal government and the corporate appeared to have made amends, and the 2 had labored collectively to develop entry to Mythos. Nevertheless, now the 2 appear destined to conflict once more.