Anthropic disables all Fable 5 and Mythos 5 model access after Trump blocks foreign access
Anthropic has suspended all access to Fable 5 and Mythos 5 after the Trump administration, through the U.S. Commerce Department, invoked national security export controls on June 12. The order blocks Anthropic from distributing the models to any foreign national, including foreign governments, companies, individuals, and foreign nationals working inside Anthropic. To comply, the company decided to removed access for all customers, not only users outside the U.S.
Anthropic said its other AI models and the Claude chatbot remain unaffected. The government has not publicly detailed the specific national security concerns, but Anthropic believes the decision may be linked to a narrow, non-universal jailbreak method targeting Fable 5, reportedly shared with officials by an unnamed source.
Fable 5 launched just a few days ago, on June 9, as a public model meant to bring some of Mythos 5’s advanced capabilities to a broader audience, while Mythos 5 remains limited to Project Glasswing partners. Anthropic disagreed that a potential jailbreak should justify recalling a commercial model, arguing that government action against unsafe AI deployments should follow a transparent, fair, and technically grounded process. Pentagon CIO Kirsten Davies publicly backed the move, framing national security as a higher priority than revenue, media attention, or pre-IPO valuation.
- AI Chatbot
- Freemium
- Proprietary
Related news
External links
- Statement on the US government directive to suspend access to Fable 5 and Mythos 5Anthropic Newsroom • Official source
- Scoop: Trump admin blocks foreign access to Anthropic's most powerful AIAxios
- Anthropic disables top-tier AI models after US order limiting foreign accessReuters
- Anthropic Blocks All Customers' Access To Fable 5 And Mythos 5Engadget
Comments
If a narrow jailbreak is the stated trigger, the open question is why pulling commercial access for all customers — including US-based ones — was the chosen remedy rather than, say, targeted restrictions or a patch window.
Maybe next time don't hype your product as a cybersecurity risk before launch.
Gee, I wonder who reported them. It couldn't be a super-jealous competitor, can it?
More AI hype headlines before their IPO 💤
Being officially blocked for security reasons after teasing that its model is "too dangerous to be released" is indeed great publicity.
Because researchers have been able to reproduce all documented Mythos findings using open-weight models (that are 100x cheaper) and got sometimes even more details just by using Opus. On "thousands" of supposed vulnerabilities found using Mythos, none of them have CSS or CVE, so the burden of proof is still on Anthropic.
And the blocking from the current administration isn't based on anything. Chinese made internet routers have also been banned for security concerns (except few companies like Cisco and Amazon, that produce most if not all of their routers there), not to mention TikTok that was supposed to a great loophole in national security. It's just damage control since Salt Typhoon hack has been discovered (and not fixed so far) and DOGE has cancelled most CISA fund.