It's weeks like this one when I become convinced that an evil boogeyman destroyed the American education system on purpose.
It's weeks like this one when I become convinced that an evil boogeyman destroyed the American education system on purpose. It's weeks like this when I'm convinced that compulsory education was made so that people wouldn't learn to think for themselves. And it's weeks like this when black guilt creeps in telling me I must vote since people died for me to vote. I can't just exit stage left at the ballot box because participation in the legislative process seems largely useless. It's not like we follow them anymore. We haven't for decades, or big tech wouldn't be skirting monopoly laws. So, you know what, I just don't want to care anymore.
What has caused this deep sense of disgust and apathy? None other than the Trump administration blocking the new Anthropic Claude Fable model. They claimed it is a security risk, but it's real political penalization for refusing to be a branch of the US military. Trump administration is hoping that the illiterate and well-entertained US public didn't notice. And most haven't, just this small, yet vocal corner of the internet.
This supposed security threat is a figment of Trump's imagination. Because every single LLM can do what this Fable model can. (Anthropic statement, June 12, 2026).
And that, dear reader, is to review code for exploits. A tester of the model requested a code review. That review found security flaws. The human cybersecurity expert developed a method through that to jailbreak the model. So, are we shutting down Coderabbit and banning the study of security exploits now? All LLMs can do this, even the open source ones you can host on your laptop. The AI did not develop the exploits intentionally or alone. And here's the part that should make your blood boil: Anthropic red-teamed Fable with the US government, the UK's AI Safety Institute, and multiple third-party organizations for thousands of hours before launch — and the government signed off. Anthropic followed all known security testing protocols, and now even their own developers who aren't US citizens can't work on the Fable model anymore (Commerce Secretary directive, full letter via Reddit) . The only reason for the current administration to call this a security threat is:
Money and control.
This will screw with the IPO of the number one competitor against this administrator's preferred military vendor (OpenAI).
Trump is a small man with a big ego telling him "No" hurt his feelings and no one hurts his feelings.
I'm totally for Anthropic taking this through the court system. It will either destroy LLM development throughout the entire country, prove that we are just cogs in the machine to be toyed with if we refuse compliance, or uphold basic American principles of not being bullied by the government. Either way, this is a pivotal moment in American history.
On Tuesday I'm breaking down exactly how AI security testing works, what a jailbreak actually is, and why this directive would shut down the entire industry if applied consistently. Subscribe below ($9/month or $60 a year)
This has a been A Geeky Production
