Anthropic, started refunding money to users! Just today, the United States banned Claude Fable 5. Anthropic immediately decided to close the model for all users and start a refund service with a deadline of June 20.
Netizens who knew the news excitedly spread the word all over the Internet.
Refund deadline: June 20
If you need to cancel your subscription, please be sure to use a computer browser to do so. This cannot be done on the mobile client. Open the corresponding notification email and click Unsubscribe according to the prompts on the page.
Note that the deadline for refunds is 11:59 pm on June 20, 2026.
In addition, Apple channel users please note: If you have previously subscribed through Apple Store Apple in-app purchase, Anthropic has no right to handle it directly. Refunds can only and must be applied for through Apple’s official support channels.
According to information online, some users have already received the email discount, but not all users can receive it. Other users need to apply manually through customer service chat.
User fryer
I recharged $220 and played for a day? !
As soon as the refund channel was opened, communities at home and abroad were excited.
A refund seems like a good thing.
But in fact, what most netizens really want is to bring Fable 5 back.
Moreover, Anthropic still has a lot of cool tricks.
Many users who paid $220 found that they could only get half of their money back.
Some netizens expressed angrily: "I spent $220 to upgrade Max 20x just to play Fable 5. But the model is gone! When I clicked on refund, I was actually told that I could only refund the price difference of the upgrade proportionally. The basic subscription was not refundable, and all access rights were taken back directly. Even downgrading back to the basic version was not allowed!"
Some people have even found that they were judged ineligible for a refund because they canceled their subscription too quickly.
And more sober netizens interpreted a layer of despair from Anthropic’s high-profile global refund:Fable 5, there should be no return in the short term.
There are still many people who want to say to Anthropic: If you don’t seek death, you won’t die.
Who stabbed Anthropic in the back?
So, who sentenced Fable 5 to death?
What triggered it all was a demonstration.
According to Axios, "another company" demonstrated a method to defeat Fable 5 to the U.S. Department of Commerce.
The Wall Street Journal says that company is Amazon.
In the demonstration, they successfully proved that although Anthropic claims to have conducted more than 1,000 hours of red team testing, as long as a specific mysterious code is entered, all security restrictions of Fable 5 can be instantly lifted and anyone can command it.
After watching the demonstration, the Washington officials were so shocked that they broke out in a cold sweat.
Anthropic learned of this and quickly launched a counterattack:
What the other party showed was not a universal jailbreak at all, it was just an extremely narrow, non-universal vulnerability! Essentially it just asks the model to read a specific code base and fix the vulnerability.
When we went to review the demo ourselves, we found only a handful of previously known, minor vulnerabilities.
What’s even more cruel is that Anthropic directly launched indiscriminate attacks, bringing the entire industry down.
"Other public models, including GPT-5.5, can expose similar problems without even needing to be jailbroken! It is completely unreasonable to block us with a border vulnerability that cannot be perfectly solved by the entire industry!"
But Washington doesn't think so. Prior to the June 9 launch, the government had privately tried to dissuade Anthropic from deploying the two models.
Sure enough, after the persuasion failed, the export control letters arrived.
Dario’s words came true: This time it’s a big deal
What’s funny is that Dario actually made his prediction come true.
They have long been proactive in seeking regulation.
On June 10, Dario also wrote: The government should have the power to prevent or ban the deployment of high-risk models.
This time, he succeeded.
A few months ago, Ultraman mocked this tepid approach, saying it was like "Anthropic built a bomb and then sold you a bulletproof bunker."
Security researcher Peter Girnus ridiculed: You market your own bombs every day, and now the Department of Commerce has officially recognized it as a bomb.
The foreign media TechCrunch’s title is the funniest: “Anthropic’s security warning may have just backfired on itself—the U.S. government unplugged its most powerful AI.”
This time, Anthropic vividly explained what it means to play with fire and burn yourself.
What really troubled Anthropic was the timing of the whole thing.
Anthropic secretly submitted an IPO application on June 1 with a valuation of about US$965 billion. It aims to go public in the fall. Annualized revenue has climbed to US$47 billion.
This was supposed to be a perfect listing story, but now the building collapsed.
On the illiquid secondary market, Anthropic's shares fell several percent in a single day.
For investors, there is a new line of content in risk assessment: the most valuable products may be disabled overnight.
Will Fable 5 be resurrected? No one knows.
This time, Anthropic accidentally lost his game.
References:
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