Less than an hour ago I received a Breaking News alert on my phone announcing, “U.S. Lifts Restrictions on Anthropic’s Most Powerful A.I. Models.”

Yay! Does the mean I get my favorite A.I tool Claude Fable 5 back? There are so many things that I both love and loathe about this latest, greatest and extremely powerful AI model.

Just weeks ago, I was sitting outside at a restaurant with a friend and we both lamented the loss of Claude Fable 5, which was snatched from consumer’s hands following an executive order banning this paired down version of Claude Mythos with guardrails.

“Claude Fable 5 was so good, and so smart!” we agreed. So much better than all of the other LLM’s that we both used regularly.

Claude went from the capabilities of a clueless high school intern to a thoughtful grad student in a short period of time.

This past April 7, 2026 Anthropic, Claude’s parent company, banned its latest model Mythos Preview from public release. Why? Because it can find vulnerabilities in every bit of software in existence.

In other words, Claude Mythos Preview is a hacking genius capable of breaking into banks, medical records, big tech (Apple, Amazon and Google accounts) the same way a locksmith can break into a school locker combination—easily.

So it was released to a few organizations and entities so they could patch their vulnerabilities in advance. The existence of Claude Mythos makes it official, computers are better at writing code than people. More on the implications of this later.

But on June 9, 2026 Anthropic released Claude Fable 5, a version of he underlying Mythos model but with safeguards. And I, along with thousands of others immediately started to use it and love it.

But then three days later, on June 12, 2026, Anthropic disabled access to Claude Fable 5 and Claude Mythos 5, after receiving a U.S. Department of Commerce export-control directive that required blocking access for foreign nationals, including employees abroad. In response Anthropic cut off access to all users, foreign and American citizens inside the U.S.

But let me back up. I wasn’t always a huge fan of Claude. Almost two years ago, I tested Claude, Perplexity and ChatGPT with the same series of prompts to see which was best at assisting me, back then Claude failed miserably and came in last place. But things change fast, exponentially fast in the realm of AI development.

I had used Claude Fable 5 to audit my professional website, and it did an outstanding job. It identified, writing inconsistencies, extra spaces around punctuation and between words and even found a page where I had the wrong meta description on a page because I’d created a new web page by copying an existing web page and forgot to update the backend. Claude Fable 5 even checked my math. I had said that I had helped a client make a 71% increase in sales (numbers from the company’s CFO which I hadn’t double checked), but Claude Fable 5 re-calculated the sales growth (from $3.5M to $6.3M) to equal 80% not 71% as I had stated.

It also produced a whole list of minor errors I needed to fix, including extra spaces between words and writing inconsistencies.

Recently, I’ve also begun experimenting with Claude Design, which was released, April 17, 2026. And months ago, I began building a new website using Claude Code, all of which I found exciting and promising.

I uploaded to Claude a short list of apps and projects I wanted Claude Code to create for me and asked it to organize and sort projects by simplest to implement to the most complicated.

On the frivolous side, I want a personal app that knows all the clothes in my closet and helps me match up outfits quickly based on my schedule and the weather. Hot summer day, navigating Costco, Trader Joes and Safeway. Bam! The perfect comfortable, on-trend outfit to stroll never-ending parking lots and stand comfortably in cart-deep, check-out lines.

On the more serious side, I want Claude Code or Claude Design to redesign my entire professional website. I want to see if it can generate more inspired design ideas for me to elevate the aesthetics of the site beyond my Squarespace template.

But I need Claude’s help prioritizing them, since my past Claude Code project, designing a complex website turned out to be far more complicated than I expected.

As a paid Claude subscriber, I’ve made a deliberate and conscious choice to upload primarily work-related data to Claude for its parsing, organizing and assistance synthesizing. I’ve used Claude to update my LinkedIn profile, my resume and professional website, all things that are public facing.

I’ve refrained from uploading large amounts of detailed private information, like medical reports and financial data.

Things I Loathe About My Favorite AI Tool

While I love Claude Fable 5’s assistance in everyday tasks, I loathe that despite being a paid subscriber my access was cut off without warning.

You can’t give me electric power tools to build a house then take them away and expect me to be happy building with only a hammer. That’s what it felt like when Claude Fable 5 went away. It was helping me work efficiently at a level far exceeding Claude Opus 4.8, the next model down.

And yet, I’m also deeply alarmed at the places where Claude is showing up with access to my private and protected data without my consent.

I’m very purposeful not to upload medical data and financial data that I don’t want out in the universe. Yet, a few day ago I received a summary from one of my healthcare providers and I instantly recognized Claude-speak in the writing.

My healthcare providers sentence began with: “The moment that landed. . .” an awkward phrasing Claude often generates but that no human writer I’ve read would ever write.

So I messaged my healthcare provider ad asked her if she was using claude.ai and if its within HIPAA or running independently. She replied that she’d never heard of Claude so she reached out to the company’s IT department who confirmed, “The name of the system that built our tool is Anthropic Claude. This tool is HIPPA compliant built into the platform. . .”

How is it possible that neither me nor my healthcare provider knew that Claude was capturing all the details of my visit?

The moral of the story. Opting in to AI usage, as a paid subscriber to an LLM doesn’t give me full control of the tools I use. Arbitrary legislation, without warning can limit my paid access.

Simultaneously, that same AI tool is being given access to my private, medical information without my consent and knowledge.

We all know data breaches are common. In fact Anthropic disclosed an accidental code leak of Claude Code on March 31, 2026, and separate unauthorized access to Claude Mythos was reported in April 2026 after the model’s limited preview release earlier that month.

For better or worse: The restriction on Claude Fable 5 was eased on June 26 for Mythos 5, and on June 30, 2026, the Commerce Department lifted the controls for both models, with Fable 5 scheduled to come back online on July 1, 2026 (tomorrow).

Perhaps I’m naive to hope that Claude Mythos was used to make the portal where my healthcare data is stored unhackable but I guess I’ll have to wait for the next data breach alert to know for sure.

AI is truly everywhere all at once. Not just in healthcare, but in our web browsers, our schools, our workplaces, our public spaces (biometric readers at airports and via AI-powered eyeglasses and dictation tools that see and record all).

Whether I love it or hate it, one thing is clear. I have no control alone over how it is used by me or on me.

But collectively, we can elect officials and rally for laws that regulate how AI is deployed and for control over the AI that is showing up everywhere, all at once.