Cyber Experts Urge White House To Reverse Ban On Anthropic’s Fable Model

Shutting down public access to advanced AI tools like Anthropic’s Fable might feel like a security win—until you realize defenders are now disarmed while attackers keep right on going.

The U.S. government’s abrupt directive to block foreign users from accessing Fable backfired instantly: Anthropic just turned it off for everyone. The model was already a tamer version of Mythos, designed with safeguards for public use. But according to Amazon researchers, those guardrails could be bypassed to launch cyberattacks—prompting the clampdown.

Here’s the problem: as 150 cybersecurity experts point out in an open letter, this move only hurts the good guys. Jack Cable, former DHS researcher and CEO of AI cybersecurity startup Corridor, puts it bluntly—defenders are now hamstrung while adversaries continue using similarly powerful AI, often outside U.S. reach.

I’ve seen how reliant modern security operations are on AI-driven vulnerability discovery. When one tool disappears overnight, it exposes a fragile dependency. Qasim Mithani of Depthfirst argues that’s exactly the lesson: don’t rely on any single model. Defenders need autonomy, control, and diverse tools to stay ahead.

And the cat’s already out of the bag. Oege de Moor of Xbow notes that GPT-5.5 has been public since April and is nearly as effective as Mythos—at a lower cost. Even open-weight models from China, like Kimi-K2.6, are matching top-tier AI in crafting exploits. Once capabilities go open-source, no government order can pull them back.

This isn’t about loosening restrictions—it’s about recognizing reality. Security teams need resilient access to advanced tools, not sudden shutdowns based on incomplete understanding. If we lock down access without closing real gaps, we’re just pretending to be secure.

Read the full breakdown to understand why AI access is becoming a frontline defense issue—and why the response needs to be smarter than a blunt kill switch.