Fable 5 is gone. The pricing change it left behind is the real story.

The most powerful AI model ever released to the public lasted about as long as a carton of milk. Three days, and it was gone.

That model was Claude Fable 5. Anthropic put it out on June 9, called it the most capable model anyone could access, and by Friday night the U.S. government had effectively pulled it off the shelf. There is a national security story here, and a genuinely weird one about who tipped off the White House and why. I'll get into both.

But the part that matters most to you is a detail you cannot even see anymore, because the model is blocked. It quietly changed the deal for every business that uses AI. And right now it is getting buried under the showdown between Anthropic and the government.

This week:

Why Fable 5 vanished in three days, and the one detail that should worry you more than the ban

How to stop flying blind on finances

Apple paying Google a billion dollars, two giant IPOs, and OpenAI buying the piece that lets AI keep working after you log off

The Deep Dive

Fable 5 is gone. The pricing change it left behind is the real story.

Here is the timeline. June 9, Anthropic releases Fable 5, the public version of its most powerful model family. Three days later, the Commerce Department uses export controls to bar the company from giving any foreign national access to it. Anthropic says the rule was so broad it had no choice but to shut the model off for everyone. So they did, on a Friday night.

Now the weird part. The government says someone found a way to jailbreak Fable 5 into doing things it was not supposed to do, asked Anthropic to fix it before release, and Anthropic declined. That someone turned out to be Amazon. Reporting says Amazon's CEO personally called the White House to flag it, and Amazon happens to be Anthropic's single biggest investor. Meanwhile Dario Amodei, the guy who has built his whole public image on being the most safety-obsessed person in AI, is the one who reportedly looked at the flaw and shipped anyway.

So one of two things is true. Either Mr. Safety is not as careful as he sells himself to be, or the jailbreak the government got so worked up about is mostly a nothing burger. Anthropic's own statement obviously points to the second one. They called it "a small number of previously known, minor vulnerabilities" and said a narrow jailbreak should not be reason to recall a model used by hundreds of millions of people.

And here is what really does not add up. This was not some surprise that showed up after launch. The government and its trusted partners had access to Mythos, the model family Fable is built on, well before any of this went public. And Anthropic knew about the jailbreak before release and shipped the model anyway. So everyone involved saw the exact thing they are now calling a national security emergency, in advance, and let it out the door. Either it was fine then and the panic is overblown, or it was dangerous then and they all signed off. It cannot be both.

To be clear, I think a government should be able to hit pause on a genuinely dangerous technology. That is a reasonable power to have, and Anthropic thinks so too. Just days before all this, Amodei published an essay arguing the government should be able to block or even reverse a model release that fails safety standards. Then the government did basically that, and Anthropic fought it. The principle is right. A Friday-night takedown with no warning, kicked off by a phone call from the company's own biggest backer, is not how you put that principle into practice. This needs to be a real collaboration between the labs and the government from the start, not a panic button somebody slaps after the fact.

Alright, completely changing gears, because this next part is the thing that actually touches your business.

Here is the nugget, the part you cannot even see anymore. Fable 5 was included on the regular Claude subscriptions, but only until June 22. After that, getting to it runs on tokens, billed by usage, on top of your flat monthly fee. I wrote a whole piece a few weeks back on why a $200 Claude subscription is the best deal in the world, because it hands you frontier AI for a flat price instead of paying 50x more on raw tokens. Fable 5 just broke that. It is the first time the best model is not part of the subscription at all. And it runs $10 per million tokens in and $50 per million out, while burning through roughly double the tokens of Opus.

That is the precedent worth watching. If every new top-tier model from here arrives token-only, regular businesses get priced out of the best AI the day it ships. You are now doing real math: is $200 a month for Opus 4.8 good enough, or is the newest model worth thousands a month in usage? For most businesses, most of the time, the answer is going to stay the subscription. But "the best AI is included" was a quiet gift, and it is starting to go away.

This is something we are watching closely at Bridge AI. The moment more models start arriving token-only, "which AI is actually worth paying extra for" stops being a fun thought experiment and starts hitting real budgets. We will be ready for that the day it lands. If you want to get ahead of it, reach out at [email protected].

From the Field

Stop flying blind on finances

Picture how most owners actually know how their business is doing. You wait for a monthly report. Your bookkeeper sends a P&L a couple weeks after the month closes, you skim it, and the rest of the time you are running on feel. Want to know what you spent on contractors last month versus this one? That is an email and a two-day wait. By the time you see a problem, it is six weeks old.

It does not have to work that way. Connect the QuickBooks connector to Claude, and your books stop being a locked filing cabinet you check once a month. You just ask. "What did we spend on contractors in May versus April?" "Which customers still owe me money, and how overdue is each one?" "Build me a cash flow view for the next 60 days based on what's outstanding." You get the answer in about ten seconds, in plain language, pulled straight from your real numbers. For the deeper stuff, the QuickBooks API lets us go further, custom reports and checks that watch for things you would never catch by hand.

The before and after is simple. A question that used to mean pinging your bookkeeper and waiting two days now takes one sentence and a few seconds. You go from checking your finances once a month to knowing where you stand any day you want.

A human still reviews anything that drives a real decision. Always. But the gap between "I wonder how we're doing" and "now I know" basically disappears.

AI News Worth Your Time

Three Stories From This Week

Apple is paying Google a billion dollars a year because it couldn't build it alone

At WWDC on June 8, Apple finally showed off the rebuilt Siri, and the headline is what's under the hood. It runs on a custom 1.2-trillion-parameter Google Gemini model, and Bloomberg reports Apple is paying Google around a billion dollars a year for it. The new Siri can actually understand your personal context across your messages, email, and photos, take multi-step actions across apps, and it gets its own standalone app. It ships with iOS 27 this fall.

My take: Sit with this for a second. Apple, the most valuable company on the planet, with effectively unlimited money and talent, and the one company that made privacy and protecting your personal data its entire identity, looked at building its own frontier AI and decided to rent Google's instead. If Apple is buying rather than building, the business owner agonizing over whether to "build an AI tool" has their answer. You don't build. You adopt the best thing that already exists and connect it to your work. The companies winning with AI right now are not the ones with the fanciest in-house thing. They are the ones who moved fast and used what was already there. READ THE BUSINESS STANDARD REPORT →

OpenAI and Anthropic both filed to go public, days apart

Within a week of each other, both of the biggest AI labs filed confidentially for an IPO. Anthropic filed June 1 at a reported $965 billion valuation, fresh off a $65 billion raise. OpenAI filed June 8, targeting somewhere between $730 and $850 billion, with some reports floating a trillion. These are the companies that build the AI most of us actually use, about to become public companies answerable to shareholders.

My take: This is a "they're here to stay" moment, and it should make you more comfortable building your business around these tools, not less. But notice the other edge of this news. Public companies have to make money, and that pressure is real. Fable 5 moving to token pricing up top is exactly what that pressure looks like in practice. Get good at squeezing value out of the affordable tier now, while the gap between $200 a month and the absolute best is still small. READ THE BLOOMBERG REPORT →

OpenAI bought the piece that lets an AI keep working after you close your laptop

On June 11, OpenAI agreed to acquire Ona, the German startup formerly known as Gitpod. Ona's whole thing is secure cloud environments where an AI agent can keep grinding on a long task, hours or even days, after you have walked away from your computer. Folded into OpenAI's Codex, it means you can hand off real work and the agent runs it inside your own cloud, where your company keeps the data, the logins, and the paper trail. Around two million developers already use it.

My take: Ignore that this one is pointed at coders. The direction is the part that matters, and it points at every business. We are moving from AI you sit and chat with to AI that goes off and does a multi-hour job on its own while you do something else. And notice what everyone is quietly fighting over: keeping it inside your own walls, with your data and your access under your control. That is the same instinct behind Apple guarding personal data up top. The winning setup is not "ship all our stuff off to someone else's computer." It is an agent that does the long work for you while your information stays yours. READ THE OPENAI ANNOUNCEMENT →

That's it for this week.

If you found this useful, forward it to someone who is trying to figure out whether the newest, most expensive AI is actually worth it for their business.

And if you want to talk about what AI could actually do for your business, not in theory, but in your workflows, with your team, on your timeline, reach out at [email protected]. That's what we're here for.