Anthropic pulls – Anthropic has taken Claude Fable 5 and Mythos 5 offline to comply with a US government export-control directive tied to national security concerns, disabling access for all customers and reigniting a broader dispute over how the administration treats the compa
The last thing customers expected to see was a sudden lock on an AI model they had only just met.
Anthropic says it has disabled access to two AI models it launched earlier this week—Claude Fable 5 and Mythos 5—to comply with an export control directive it received Friday afternoon from the US government. citing national security concerns. The company says the order targeted access by “any foreign national. whether inside or outside the United States. including foreign national Anthropic employees.” To ensure compliance. Anthropic removed access for all of its customers. not just some subset.
This marks the latest flare-up in a growing conflict between Anthropic and the Trump administration. Earlier this year. Trump’s Department of Defense labeled Anthropic a “supply chain risk” after the Claude-maker pushed for what it described as red lines over how US military users could deploy its technology. That designation effectively barred government agencies and contractors from using Anthropic’s systems. Anthropic responded by filing lawsuits against the Trump administration.
On Tuesday. Anthropic publicly released Claude Fable 5. a version of the company’s Mythos AI model with safeguards meant to prevent it from answering questions about cybersecurity. biology. and chemistry. Before the public release, Anthropic said its Mythos Preview AI model had a limited rollout in April.
Anthropic said the earlier Mythos Preview rollout was designed to give companies and organizations a chance to use its cybersecurity capabilities to strengthen defenses—while also addressing concerns that the technology could be exploited to build powerful hacking tools. In the same vein. Anthropic says it conducted work “in collaboration with the US government” before Claude Fable 5’s public release.
Then came the letter.
In a blog post on Friday, Anthropic said it received a letter from the US government at 5:21pm ET. The company wrote that the letter “did not provide specific details of its national security concern.”
Anthropic says it understands the government believes it had discovered a method for bypassing, or “jailbreaking,” Fable 5. The company added that it reviewed a demonstration of the technique being used to identify a small number of previously known. minor vulnerabilities. Anthropic says the vulnerabilities “all appear relatively simple. ” and it found that other publicly available models can discover them without needing a bypass.
Anthropic also argued that its safeguards reduce the likelihood of misuse and that the jailbreak the government found for Claude Fable 5 was narrow. In its view, that narrowness meant it would not have made an attacker meaningfully more dangerous than they would have been with another AI model.
“To date. the government has only given us verbal evidence of a potential narrow. non-universal jailbreak. which essentially consists of asking the model to read a specific codebase and fix any software flaws. ” Anthropic wrote in the blog post. “Our understanding is that one potential jailbreak was shared with the government.”.
The company’s deeper point is that the government action doesn’t match the process it says it wants. Anthropic said earlier this week in a policy essay by CEO Dario Amodei that it supports a fair. structured. and transparent government process to block unsafe AI releases. In the blog post Friday, Anthropic argued that the step it took “does not adhere to those principles.”.
At the White House and the US Commerce Department, spokespeople did not immediately respond to a request for comment.
Even with Anthropic insisting the identified bypass was narrow. the timing is difficult to ignore: the model was released publicly on Tuesday. then taken offline after a directive arrived Friday afternoon—first constrained to specific categories of users. then broadened into a full customer disablement to meet the order’s requirements. For a company trying to move from controlled previews into wider use. the disruption lands as the newest reminder that—when national security is cited—the rules can change faster than deployment plans can absorb.
Anthropic Claude Fable 5 Mythos 5 US export control AI models offline cybersecurity safeguards jailbreaking national security Dario Amodei Trump administration
So they just turned it off for everyone? That’s wild.
I don’t get it, like why would an “export control” shut it down inside the US too? Sounds like politics more than security. Also people were just starting to test it.
Wait, I thought the government was the one testing AI for national security, not ordering it offline like a TV channel. Export control usually means overseas stuff, so “foreign national employees” is confusing. Isn’t that just… everyone with dual citizenship? Kinda sounds like a blanket punishment.
This is the same thing as when they say “safeguards” but then someone always figures it out anyway. If Claude Fable 5 was supposed to avoid cybersecurity/biology/chem questions, why not just tighten that part? Turning it off for all customers seems like they’re trying to make a point in the Trump vs Anthropic drama.