While the rest of the country celebrated the United States’ first World Cup win and the New York Knicks’ championship, Anthropic spent the weekend fighting the Trump administration over its latest model release. At 5:21 pm on Friday, the company received a US export control directive to suspend access to its Mythos 5 and Fable 5 AI models by “any foreign national” inside or outside the US, “including foreign national Anthropic employees.” Anthropic determined that the only way that was possible was to completely disable the products it had promoted the previous week — and traveled to Washington, D.C., in hopes of changing President Donald Trump’s mind. Now, in the coming days, the US government may dramatically change the trajectory of the entire industry, dealing a major blow to US AI companies.
Cloud Mythos 5 and Fable 5 are built on the same foundation as Anthropic’s Mythos Preview, which Anthropic has deemed too dangerous to release publicly. (The company’s warnings could be seen as genuine concern or more promotion for their own model – or both.) The Mythos 5 was made available to a select group of government agencies and companies, while the Fable 5, which included additional security measures, was deemed “safe for general use”. But when a report indicated that those guardrails had failed, Anthropic’s dire warnings about Mythos falling into the wrong hands came back to haunt him.
A source familiar with the situation, who participated in talks between Anthropic and the Trump administration, said the administration called AI Labs around 1 p.m. ET on Friday and gave the company a 90-minute ultimatum to cut off access to Mythos 5 and Fable 5. If this does not happen, the government will impose export controls on Anthropic with the authority of the US Commerce Department.
The source said Anthropic officials were talking to the White House within 15 minutes of that first call, confirming that CEO Dario Amodei joined the discussion about an hour and 15 minutes after that initial call. Amodei spoke directly with U.S. Treasury Secretary Scott Besant, Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross, in some cases more than once, the source confirmed.
Anthropic wrote in a release Friday that the company believes the government “believes it has become aware of a method to bypass, or ‘jailbreak,’ Fable 5.” However, rather than an existential threat, Anthropic said the jailbreak in question was a “potentially narrow, non-universal” one that was “shared with the government” by an entity that the company declined to name. Additionally, Anthropic stated that the behavior was not unique to Fable 5. “We have reviewed a report, which we believe to be the basis for the government’s guidance, and have confirmed that the level of capability demonstrated there is consistent with what is widely available from other models (including OpenAI’s GPT-5.5),” Anthropic wrote.
Semaphore Citing a source familiar with the situation, the report said the uproar began because the US government was concerned that a China-linked group had gained access to the technology. But the source said the China rumors go back several weeks, referring to a large global telecommunications company that was initially approved for inclusion in access to the Mythos preview, and when the U.S. government shared its concerns, Anthropic immediately revoked access.
An X post by David Sachs, the former US government AI and crypto mogul who stepped down in March, also did not mention China. However, Sachs noted the unnamed entity that exposed the issue to the government, calling it “a highly regarded trusted partner of both Anthropic and the USG that was testing Fable.” [which] Come forward with breaking those railings.”
Some reports point to Amazon CEO Andy Jassy as the person who raised concerns to the US government after the Fable 5 was red-teamed by Amazon researchers. This conclusion is in contradiction with some independent red-teamers, who have said that they were impressed by the level of security.
The source familiar with the talks said Amazon research was explicitly mentioned in talks with the US government. The person said Anthropic had gained access to the paper within days of Friday’s export control directive and had been reaching out repeatedly to discuss it with Amazon researchers since then.
Everything in that paper, the source said, can be achieved by OpenAI’s GPT-5.5.
Anthropic spent the weekend making good with the Trump administration, starting with virtual meetings and then flying staff to DC, including Anthropic’s head of security measures, Dave Orr; Logan Graham, who runs its Frontier Raid team and has led work on Project Glasswing; and Nicholas Carlini, a leading frontier developer and cybersecurity researcher. Axios Citing a source familiar with Trump administration thinking, the report said the company has repeatedly made missteps in its communications with the administration and has “not done a good job of trying to talk to the administration and appreciate ideological differences.” For Anthropic, the timing couldn’t have been worse: The company had relied on Mythos to help it recover from months of high-profile clashes with the US Department of Defense.
The source familiar with the talks said that while Anthropic had given the administration a pre-deployment briefing about Fable 5, and the US Commerce Department had conducted pre-deployment testing, no concerns were shared at the time. The source said Anthropic had been working closely with government agencies since the release of the Mythos preview.
The Trump administration initially took a lax approach to AI safety — but after Mythos, it has become more ambivalent, even as it worries about the danger of losing the AI race to China. Now, leading cybersecurity leaders warn that bypassing Mythos 5 and Fable 5 could give China a significant AI advantage. Trump’s move has prompted international calls for alternatives to American AI systems, while effectively shelving a new flagship model from a major American AI company.
A public letter from technical and cybersecurity officials on Sunday called for restrictions on the Fable 5 to be repealed. “We do not all agree that AI regulation is the right way to move forward,” the letter said. The letter says that if regulations are going to be in place at all, they should be rooted in “scientific assessments developed with input from industry and academia.”
Alex Stamos, Chief Product Officer of Corridor, said: The Verge They organized the public letter because countless vulnerabilities written in various coding languages over the past decade require AI to be fixed before bad actors can find them. “We’re in a race, and I think policymakers don’t understand that,” Stamos said. “There’s this strange arrogance, this idea that US laboratories are so far ahead of our competitors that will always be true, because of that it’s really important to restrict access. I think that’s silly. If the laboratories are ahead, it’s only a matter of a few months. And you can see that in the open assessments. The state-of-the-art models are only six months ahead of the Chinese models – and these are the models that we know about.”
The public letter said that although Anthropic’s Mythos-class models are adept at finding cybersecurity vulnerabilities and exploiting exploits, they are not “typically good” at these tasks and that the Fable 5’s security measures were “so invasive that it became a source of humor in the cyber community on launch day.” Stamos told The Verge that “There’s really an exaggeration of Mythos’ capabilities. Anthropic itself is somewhat responsible for this, obviously… Mythos is great, but the real turning point was really last year.”
Stamos said the industry is filled with backup contracts being signed with non-U.S. companies and open-weight models being deployed on alternative hardware arrangements as last weekend made political risk more a part of companies’ business plans than ever before.
“They’re laughing at us in Beijing right now,” Stamos said. “A champion of America is being kneecapped by the US government while we are in a race with the Chinese. This is incredibly stupid. That’s why I wrote the letter, and I think that’s why so many people signed it.”
Ben Van Roo, co-founder and CEO of Legion Intelligence, a system of agents for the national security community, told The Verge That “the directive ‘No foreign national should use this model’ is the most impossible thing to enforce.” He added, “When I read it for the first time, my whole… [network of] The AI community was exploding.”
To make matters more urgent, OpenAI, Google, and Microsoft have all come out with their own comparable products to Anthropic’s Mythos, and are making similar claims about their effectiveness and risks. If the Trump administration bans Anthropic’s advanced cybersecurity models, it could make a case for banning its competitors’ models as well. This could inspire AI industry leaders to band together and help Anthropic, or with its battle with the Pentagon over autonomous weapons, establish itself as a safer and more compliant alternative.
Even as the Trump administration tries to free tech companies from regulatory constraints, the Anthropic Order could impose a dramatic ban on powerful AI models — depending on how the next few days play out.
Legion Intelligence’s Van Roo called it “uncharted territory” in the regulatory setting, and said he doesn’t think it’s the last time something like this will happen.
We have also entered the era of AI populism, when large numbers of people are taking action against outside influence and the concentration of power at the top of the AI industry through data center protests, pledges to abandon the use of AI chatbots, lawsuits over wrongful deaths, and even attempted attacks on AI company CEOs. Van Roo says the Trump administration’s recent moves against Anthropic “could potentially create more fear and concerns for the wrong reasons.”
The source familiar with the talks described the weekend talks as constructive, with some members of the administration acknowledging that imposing export controls on model providers is not ideal, since competitors with similar products could find themselves under similar restrictions — and since the U.S. government is currently exploring a program that would encourage exports of American AI systems.
Monday’s talks ended without any solution so far.
As Anthropic continues negotiations with the US government, there is little chance that the company’s other myriad issues with the Pentagon will not come up – namely, the ongoing battle between Anthropic and the Department of Defense over acceptable use policies for Anthropic’s technology by the US military.
“This is new and we’ve never had anything this drastic before, and it has some real implications in terms of how we enforce access to powerful models,” Van Roo said. “Who will use this new technology that is outstripping our own ability to regulate it?”
<a href