Anthropic launched its two newest AI models on Tuesday. By Friday, the government told it to shut them off.
By Monday, the firm was in Washington trying to fix it.
Friday started normal. Then the phone rang at 1 p.m.
The government gave Anthropic an order that cited security. It said to cut off the two new models for any foreign user.
That meant users inside and outside the U.S.
The models are called Fable 5 and Mythos 5. A formal letter followed a few hours later.
That letter landed around 5:30 that evening. By then the models were already going dark.
Anthropic didn't want to break the rules. So it switched the models off for everyone, even paying users.
The timing stung. Anthropic had shown off both models just three days before.
It says it worked with the government to test them. It even got the green light to launch.
Three days later, that approval didn't seem to matter.
Market Briefs keeps you ahead of stories like this every morning in five minutes, and joining gets you a free investing masterclass.
This isn't a one-off. Back in March, the Pentagon called Anthropic a supply chain risk.
It then barred defense firms from using its tech. Anthropic is suing to undo that.
The defense chief piled on. He posted that every passing day proves the move was right.
The two sides also disagree on how big the problem is. The government flagged a security risk.
Anthropic calls it a narrow flaw.
The worry is a so-called jailbreak. That's when a user sneaks past a safety guardrail.
In this case, they could ask the model to scan code and fix its flaws.
The new models grew out of an earlier tool that was good at spotting software holes. Anthropic had only shared it with a small group of firms.
Fable 5 went out to paying customers, while Mythos 5 stayed locked to a select few.
New safeguards were meant to block risky uses, like cybersecurity and biology.
Anthropic says one narrow flaw is no reason to pull a tool used by hundreds of millions of people. It argues the same rule would freeze new AI across the whole industry.
For investors, the lesson is bigger than one firm. AI startups are racing toward the stock market at huge values.
The AI sector just learned that Washington can flip the switch overnight.
That risk hangs over the wider AI boom, not just one company. Rules can shift with one phone call, and that uncertainty has a price.
Monday's meeting is the first test. Can the two sides settle this fast?
How it ends could set the tone for how hard the government leans on AI. Other AI firms will be watching just as closely.
For now, one phone call took two flagship models offline.
Read what 350,000+ investors read each morning - it costs nothing to join, and a free 45-minute investing masterclass comes with it.
How To Use ChatGPT & Claude To Research Stocks Like A Pro
Save my seat →You’re in!
Make sure to check your inbox for the first email.
Before you go
We’re hosting our first ever livestream on how investors can use ChatGPT & Claude to invest like a pro. It’s on 11/17 and is hosted by our CEO Jaspreet Singh.
Click here to join →
