Pour yourself a massive cup of coffee and sit down, because the tech world just had its most dramatic weekend of the year.
Remember how Anthropic spent months telling the world its newest AI was far too dangerous to release freely, well guess what? The government took notes.
Here’s the absolute madness that went down.
The US Commerce Department dropped a nuclear-level export control directive late on Friday, June 13, ordering Anthropic to immediately cut off access to its top-tier models, Claude Fable 5 and Claude Mythos 5, for all foreign nationals.
And we mean everyone, whether they are living abroad or sitting right inside the United States on a visa. It even bans Anthropic's own foreign-born employees from using the tech they helped build!
Because it’s practically impossible to check everyone's passport before they log into an API, Anthropic responded in the only way they could: they abruptly killed global access to both models overnight.
If you log into Claude right now, it literally reads "Claude Fable 5 is currently unavailable."
To be clear, Fable 5 had only been publicly available for three days before the plug got pulled. Talk about a short but iconic run.
So, what allegedly caused this total panic?
The Trump administration's official concern was that someone had successfully cracked Fable 5's safety guardrails, opening a backdoor for users to extract highly restricted information.
To catch you up on the tech setup:
Claude Mythos 5 is the raw, ultra-powerful base model that was already heavily restricted to just 50 vetted organizations for defensive cybersecurity work.
Claude Fable 5 was Anthropic's public-facing version of Mythos, which was fitted with intense classifiers to block risky outputs in sensitive areas like cybersecurity and biology.
Anthropic reviewed the government's top-secret demonstration and immediately pushed back with receipts. They’re calling this whole thing a massive "misunderstanding."
According to Anthropic, the government only provided verbal evidence of a "potential narrow, non-universal jailbreak." The terrifying hack in question? It essentially just amounted to asking the model to look at a codebase and spot routine software bugs. Not exactly the stuff of Hollywood spy thrillers!
Anthropic pointed out that these minor flaws were things they already knew about, and frankly, any public model on the market could find them without any special workarounds. In fact, the company argued that this exact "level of capability" is already widely available in other public models, including OpenAI's GPT-5.5.
The kicker? Anthropic says this kind of bug-hunting is used routinely every single day by cybersecurity professionals for defensive purposes. It’s literally their job! So why were they tergetted?? I mean, it does look fishy.
But wait, weren't Anthropic's safeguards supposed to be totally rock-solid? That’s definitely what they thought!
Before launching, Anthropic spent thousands of hours running Fable 5 through intense stress-testing with the US government, the UK AI Security Institute, and independent red-teamers. Nobody found a broad, catch-all bypass. In fact, users had actually been complaining that the guardrails were too restrictive.
And yet, here we are.
This whole shutdown landed just ten days after President Trump signed a brand-new executive order creating a government framework to review national security risks of powerful AI models. White House AI adviser David Sacks took to X to spill the real behind-the-scenes tea. Sacks revealed that the government actually issued the export control "reluctantly."
The real gossip? The crackdown only happened after Anthropic CEO Dario Amodei personally refused to patch the specific vulnerability or take the model down voluntarily. Sacks noted that the administration actually wants Anthropic to just fix the security issue so they can lift the restrictions and bring Fable 5 back to life.
🍿 Sam Altman Is Somewhere Grinning??
Anthropic has officially stated that while they strongly disagree with the call, they are complying with the law and actively working to restore access as soon as possible. But they didn't go quietly into the night.
They warned that if a government can just arbitrarily shut down commercial AI deployments without a transparent, technically grounded process, the entire frontier AI industry will grind to an absolute halt.
Meanwhile, OpenAI CEO Sam Altman is probably somewhere sipping tea and laughing. Back in April, Altman told podcaster Ashlee Vance that Anthropic's safety-first marketing was textbook fear-mongering.
"We have built a bomb. We were about to drop it on your head. We will sell you a bomb shelter for $100 million," Altman joked at the time.
He might not have predicted an actual federal shutdown, but that petty observation has aged uncomfortably well.
The ultimate irony here is just pure chef's kiss. Anthropic built its entire public identity on being the hyper-cautious, safety-first AI lab, constantly warning the world that their technology was terrifyingly dangerous. But when you scream the loudest about how dangerous your tech is, the people in charge of national security are eventually going to believe you.
Turns out, those loud safety warnings have a funny way of opening the most inconvenient doors.
Oh, and don’t forget to meet us on YouTube so you can get our full take on this later today!
