76 Cybersecurity Experts Are Asking The U.S. To Lift Its Ban On Anthropic's Top AI

The U.S. pulled its most powerful AI to keep it from the wrong hands. Now 76 security experts say the move helps attackers and leaves defenders exposed.

The Letter

On Friday, the government told Anthropic to limit two models, Fable and Mythos. It cited national security but did not say exactly why.

So Anthropic cut off access for everyone, everywhere. The pushback came fast.

It took the form of an open letter signed by 76 experts. One is Alex Stamos, the former head of security at Facebook.

The list runs deep. It also includes Bugcrowd founder Casey Ellis and SocialProof chief Rachel Tobac.

Their point is blunt. The order took the best tools from defenders, the people who use them to find and fix flaws in their own code.

We track how AI policy moves markets, and which firms it helps, in Market Briefs in about five minutes a morning, plus a free investing masterclass when you join.

Why Defenders Are Upset

Here is the catch with security AI. The same model that finds a flaw to attack is the one that finds it so you can fix it.

Take it from one side and you take it from both. It is a bit like banning the best lock pick in town. The burglars keep their tools. The locksmiths lose theirs.

Good guys use these tools to harden their own systems. Bad guys do not need permission. So a ban mostly ties the hands of the good guys.

This fight has some history. Mythos first launched in April to about 50 firms, then grew to roughly 150 groups in 15 countries.

Last week Anthropic put out Fable, a public version with tight limits. Those limits block hacking work, plus germs and chemicals that could be turned into weapons.

The Disputed Paper

Anthropic said the order may trace back to one report. It claimed someone found a way to "jailbreak" Fable.

A jailbreak means tricking the model past its safety limits to unlock its full power. One signer, expert Katie Moussouris, says she read the private paper behind the scare.

Her take is that it was not a real jailbreak. The team just asked the model to fix code with known and planted bugs.

She calls that normal defense work, not a "guardrail bypass." The letter says the same trick works on other models, like OpenAI's GPT-5.5 and China's Kimi.

In other words, the team argues this is not unique to Anthropic at all.

Worth Noting

The fight is really about who gets to use powerful AI, and on what terms. The letter asks for clear rules set through a fair, public process.

It does not want a sudden order with no reason given. For investors, this is the kind of shift smart money watches closely.

One policy call in AI can move the market fast, and that can ripple straight into AI stocks.

Want to see where the AI money heads next? Join Market Briefs and you will get a 45-minute investing course thrown in.