They Gave Us the Most Powerful AI Ever Built. Then the Government Took It Back Three Days Later.
Fable 5 launched June 9th. By June 12th it was gone. Here's the full story of what happened, who was behind it, and why it matters for anyone building with AI.
So basically, this was supposed to be an article about how incredible Fable 5 is. And it is incredible. But then Friday happened, and now it’s a different article.
Let me catch you up.
The Model
On June 9th, Anthropic released Claude Fable 5. It’s their first public release from the Mythos model family, which is the most capable AI architecture they’ve ever built. We’re talking a one million token context window, 95% on SWE-Bench Verified (the gold standard for real software engineering tasks), and the kind of performance that genuinely changes what a one or two-person operation can do.
I had a close friend put it to work on a massive long-term project of his. The kind of project where just tracking everything is its own full-time job. He fed the whole thing to Fable 5. It burned through his entire five-hour Max plan limit and ran up $75 in overage on top of that. The output it came back with was actually larger and more developed than the project he put in.
That’s not a tool. That’s a collaborator. And it was available to anyone with a subscription.
For three days.
What Happened Friday
On June 12th, at 5:21 PM Eastern, the US Commerce Department issued an export control directive ordering Anthropic to immediately suspend access to both Fable 5 and Mythos 5, citing national security concerns. Anthropic complied within hours.
The order prohibited access by any foreign national, anywhere in the world, including Anthropic’s own foreign national employees working inside the United States. Because Anthropic has no real-time way to filter users by nationality, they did the only thing the order left room for: they shut both models off for every customer on the planet. US citizens included. Everyone.
Anthropic’s public statement described it as a “misunderstanding” and said they’re working to restore access. They noted the government’s letter didn’t even provide specific details about what the national security concern actually was. Their understanding is that the government believes a jailbreak method exists for Fable 5.
So who found this jailbreak and brought it to the government?
Amazon.
The Amazon Part
Here’s where it gets genuinely surreal.
Amazon is Anthropic’s largest cloud partner. They’ve committed billions of dollars to the relationship. Fable 5 literally launched on AWS the same day it launched publicly. These two companies are deeply, commercially intertwined.
According to the Wall Street Journal and confirmed by multiple sources, Amazon researchers used a series of prompts to bypass Fable’s safety guardrails and access portions of the underlying Mythos model. Then Amazon CEO Andy Jassy personally called Trump administration officials, including Treasury Secretary Scott Bessent, to share the report. The White House decided to act. Trump signed off. By Friday evening, Fable 5 was gone.
To be clear about what didn’t happen here: Amazon did not bring this to Anthropic first. They didn’t follow coordinated disclosure, which is the standard, expected practice in cybersecurity research where you notify the affected company privately and give them time to respond before going public or going to regulators. They went straight to the White House.
Anthropic’s position is that the vulnerabilities identified were already known and relatively minor. An independent cybersecurity expert who reviewed the report called it a “regulatory overreaction” and said it was “harmful to defense.”
Amazon is also an Anthropic investor. And a competitor, with their own AI systems in development.
I’ll let you sit with that combination for a second.
The Free Market Party Pulled the Plug
Here’s the part that’s genuinely hard to wrap your head around.
The current administration has built its entire brand around deregulation, free market principles, and getting government out of the way of American business and innovation. That’s the pitch. That’s the governing philosophy they campaign on.
And then they used a federal export control directive to kill the most advanced publicly available AI model ever released, three days after launch, because a competitor reported a jailbreak. Without providing specific details of the concern to the company being shut down.
This isn’t a left versus right issue. This is a consistency issue. Either you believe in free markets and let American companies compete and innovate, or you don’t. Pulling the plug on a US company’s flagship product, at the request of a meeting between a competing company’s CEO and a cabinet secretary, is not what a free market looks like.
It’s also worth noting that this doesn’t happen in a vacuum. The Department of Defense classified Anthropic as a “supply chain risk” back in March, in a dispute that reportedly centered on Anthropic’s refusal to make Claude available for mass domestic surveillance and fully autonomous weapon systems without restrictions. Anthropic has said publicly they consider that classification legally untenable and intend to challenge it in court.
So the backdrop here isn’t just a jailbreak. It’s a company that kept saying no to certain government requests, and a government that has now found ways to keep applying pressure.
The OpenAI Angle
And then there’s this.
While all of this is happening to Anthropic, OpenAI’s president Greg Brockman and his wife donated a combined $25 million to MAGA Inc., the primary super PAC supporting the Trump political operation. The donation arrived in September 2025 and made up nearly a quarter of the super PAC’s entire fundraising haul for the back half of that year.
Brockman said publicly that the donations reflected support for policies that advance American innovation and constructive dialogue between government and the technology sector.
I’m not here to tell you what that means. But when one AI company’s president is among the biggest donors to the current administration’s political operation, and a competing AI company just had its flagship model shut down by that same administration three days after launch, it’s a set of facts worth knowing.
What It Means for Builders
I build automation systems and platforms for small businesses. My whole thing is that a small, skilled, adaptable operation can now punch way above its weight class. Tools like Fable 5 are a massive part of why that’s true.
When a model like that gets pulled not because it failed, not because it was unsafe, but because of political and competitive pressure between companies worth hundreds of billions of dollars, the people who lose aren’t the corporations. It’s the independent builders. It’s the small businesses who would have benefited. It’s the people who can’t afford to absorb disruptions like this the way a large enterprise can.
Anthropic says they’re working to restore access and believe it’s a misunderstanding. I hope that’s true and I hope it happens fast. Because what Fable 5 represented, before Friday, was genuinely exciting.
The capability is there. The tools are real. The question now is who actually controls access to them, and why.