June 16, 2026
Patch me if you can
Feds freaked over Fable 5 after simple 'fix this code' prompt, not jailbreak
Turns out the so-called AI ‘jailbreak’ was just ‘fix this code’ — and commenters smell a political hit job
TLDR: A reported federal crackdown on Anthropic’s AI models may have been triggered not by a dramatic hack, but by the plain request “fix this code.” Commenters are split between calling it political theater and saying it proves AI safety rules fall apart when normal users ask obvious questions.
The internet has officially latched onto the most absurd part of this story: the scary prompt that helped trigger a federal crackdown on Anthropic’s top AI models may have been three hilariously ordinary words: “fix this code.” That’s according to cybersecurity veteran Katie Moussouris, who says this wasn’t some sneaky trick or movie-style hack at all. Researchers reportedly asked the AI to review buggy software, got refused, then simply asked it to fix the code instead — and it did. Cue the collective double take.
In the comments, people were not buying the official panic. The hottest take? That the government didn’t “freak out” so much as grab an excuse. One commenter called it a “retaliatory shakedown,” while another argued the real fear wasn’t cyberattacks, but that others might use these tools to improve their defenses. In other words: the drama isn’t just about what the AI did, but who gets to use it.
Others aimed their fire at Anthropic itself, saying this whole mess exposes a basic contradiction: you can’t market a model as dangerously powerful and then act shocked when simple wording gets around its limits. And yes, the jokes arrived right on schedule. Moussouris’ line about a ’90s T-shirt reading “fix this code” on the front and “this shirt is a munition” on the back became instant meme fuel. Community verdict: if this is a national security crisis, it’s also a punchline.
Key Points
- •The US government issued an export control directive suspending foreign-national access to Anthropic’s Fable 5 and Mythos 5 models, and Anthropic disabled both models for all customers to comply.
- •Katie Moussouris said the research that prompted the restriction involved a simple prompt—“fix this code”—rather than a jailbreak or guardrail bypass.
- •Outside researchers reportedly tested Anthropic models with open-source code containing known CVEs and newly written intentionally vulnerable code, then asked the systems to review and fix it.
- •Moussouris said defensive cybersecurity activities such as finding vulnerabilities, patching them, and validating fixes were exempted under changes tied to the Wassenaar Arrangement.
- •More than 100 cybersecurity leaders signed an open letter urging the Trump administration to reverse the restrictions and restore access to the models for cybersecurity firms.