Fable 5 is banned, and Anthropic has started issuing refunds.

Anthropic has started refunding users!

Just today, the United States banned Claude Fable 5.

Anthropic immediately decided to shut down the model for all users and initiate a refund service, with the deadline set for June 20th.

Netizens who learned this news excitedly spread the word across the internet.

Refund deadline: June 20th

If you need to cancel your subscription, please use a computer browser as the mobile app cannot handle this. Open the corresponding notification email and click to cancel the subscription as prompted on the page.

Note that the refund deadline is 11:59 PM on June 20, 2026.

Additionally, users who subscribed through the Apple Store: If you previously subscribed via an in-app purchase on the Apple Store, Anthropic has no direct authority to process the refund. You must apply for a refund through Apple's official support channel.

According to online information, some users have already received email discounts, but not all users will receive them. Other users need to manually apply through customer service chat.

Users are in an uproar: I spent $220 and only used it for one day?!

As soon as the refund channel opened, both domestic and overseas communities were in an uproar.

Refunds may seem like a good thing.

But in fact, most netizens really want Fable 5 to come back.

Moreover, Anthropic has quite a few questionable actions.

Many users who paid $220 found that they could only get half of their money back.

One netizen angrily said, "I specifically spent $220 to upgrade to Max 20x just to use Fable 5. But now the model is gone! When I clicked to refund, they told me I could only get a proportional refund for the upgrade difference, and the basic subscription won't be refunded. Moreover, they directly revoked all access rights, and I can't even downgrade to the basic version!"

Some even found that they were deemed ineligible for a refund because they canceled their subscription too quickly.

More rational netizens interpreted a desperate signal from Anthropic's large - scale global refund: Fable 5 probably won't be back in the short term.

Many people also want to tell Anthropic: You'll get yourself into trouble if you keep doing stupid things.

Who stabbed Anthropic in the back?

So, who exactly sentenced Fable 5 to death?

It all started with a demonstration.

According to Axios, "another company" demonstrated a way to break through Fable 5 to the US Department of Commerce.

The Wall Street Journal reported that the company was Amazon.

In the demonstration, they successfully proved that although Anthropic claimed to have conducted over 1000 hours of red - team testing, just by inputting a specific mysterious code, all the security restrictions of Fable 5 could be instantly lifted, allowing it to be controlled at will.

Washington officials who watched the demonstration were shocked into a cold sweat on the spot.

Upon learning this, Anthropic quickly launched a counter - attack:

What the other side showed was not a general jailbreak at all. It was just an extremely narrow and non - general vulnerability! In essence, it was just making the model read a specific code library and fix the vulnerability.

When we reviewed the demonstration ourselves, we only found a small number of previously known and minor vulnerabilities.

Even more aggressively, Anthropic launched an all - out attack, dragging the entire industry into the mess.

"Other public models, including GPT - 5.5, can expose similar problems without even a jailbreak! It's completely unreasonable to ban us because of a boundary vulnerability that the entire industry can't perfectly solve!"

But Washington didn't think so. Before the release on June 9th, the government had privately tried hard to dissuade Anthropic from deploying these two models.

Sure enough, after the persuasion failed, the export control letter arrived.

Dario's prediction came true: This time, it's a big mess

Interestingly, Dario's prediction actually came true.

For a long time, they have been actively seeking regulation.

On June 10th, Dario also wrote: The government should have the right to prevent or ban the deployment of high - risk models.

This time, he succeeded.

A few months ago, Altman mocked this kind of pretentious approach, saying it was like "Anthropic made a bomb and then sold you a bulletproof bunker."

Security researcher Peter Girnus mocked, "You've been marketing that you make bombs every day, and now the Department of Commerce has officially recognized it as a bomb."

The title of foreign media TechCrunch is the funniest: "Anthropic's security warnings may have just backfired on itself - the US government pulled the plug on its most powerful AI."

This time, Anthropic vividly demonstrated what it means to play with fire and get burned.

What really hurt Anthropic was the timing of the whole thing.

Anthropic secretly submitted an IPO application on June 1st with a valuation of about $965 billion, aiming for a fall listing, and its annualized revenue has climbed to $47 billion.

This should have been a perfect listing story, but now it's all in ruins.

On the secondary market with poor liquidity, Anthropic's stock price dropped by several percentage points in one day.

For investors, there is now a new item in the risk assessment: the most valuable product may be disabled overnight.

Will Fable 5 be resurrected? No one knows.

This time, Anthropic went too far.

Reference materials:

https://x.com/robinebers/status/2065751990228136427

https://x.com/kimmonismus/status/2065774154968711429

https://x.com/gothburz/status/2065601302705398034

https://deaination.substack.com/p/directionally-bad-issue-005

This article is from the WeChat official account "New Intelligence Yuan", author: ASI Revelation. It is published by 36Kr with authorization.