June 15, 2026

3 Minute Read or Listen ->

🔱 Sign up for a free demo today!

See how the latest AI and digital advocacy tools can transform your work.

THIS WEEK'S EDITION

 The Fable Shutdown And What It Means For Public Affairs

On Tuesday, Anthropic released the most powerful AI model it's ever put in public hands. By Friday evening, the U.S. government had switched it off — worldwide, for every customer, including Anthropic's own employees. By the time you read this, Fable may be back on, or the dispute may have spiraled to implicate your other AI tools. But regardless of any potential resolution or escalation, Friday's events mark a critical turning point in AI regulation, and the truth is that no one knows what will happen next. So in today's short read, we explain what happened, and show you how to protect your AI workflows — regardless of what tomorrow brings. 

What Actually Happened

The model was four days old. Then a letter arrived. Anthropic launched Claude Fable 5 on Tuesday. On Friday at 5:21 p.m. Eastern, an emergency export-control directive from the Commerce Department ordered the company to cut off access for any foreign national — anywhere in the world. Unable to comply selectively, Anthropic shut Fable 5 down for every customer on the planet. It is the first time the federal government has reached in and taken a publicly available AI model off the market. The government cited a national security concern over a jailbreak; Anthropic disputes the severity, calling it trivial and noting that rival models can already do the same thing. 

The Government's Logic Doesn't Stop At Anthropic

Here's the part that should worry everyone: the "dangerous" capability the government cited is something ChatGPT already does on request — no jailbreak, no trickery. If that's the line, every frontier model is standing on the wrong side of it. Translation: ChatGPT, Gemini, and the rest of the frontier AI models sit under the same authority. Whichever of those models your organization runs today, either directly, or through your vendors, a letter on a Friday afternoon could take them offline — with no warning, and no appeal.

Your Hedge: Vendors Who Run AI Locally

There's a practical hedge, and it's gone from hobbyist curiosity to boardroom strategy in about a year: open models that companies download and run on their own infrastructure — Meta's Llama, Google's Gemma, and Nvidia's Nemotron are just a few examples. These are no longer toys — the best open models now sit only about four months behind the frontier models. And once the models are on a private server, the government's potential reach into your tech stack just became science fiction. It's the difference between renting your apartment and owning it. Rent, and the landlord sets the rules — and can change the locks. Own the weights, and the model lives on your server, where a Friday-afternoon letter can't reach it.

Local and Better

Open models are not just subject to less regulatory risk — when properly trained, they can actually perform better than frontier models. Because when you hold the actual weights (the model's brain), you can fine-tune the model — literally re-weighting it — on your own data. A smaller open model tuned on your work can outperform a giant general-purpose model on your specific task, because it has been trained for your specific work. No closed frontier model will let you customize that deeply.

A word of caution before you rip out your GPT and Claude subscriptions: running and fine-tuning models yourself takes real engineering talent and serious hardware. GPUs are expensive, and a local model in production needs people to keep it fed, monitored, and updated. For most shops the realistic path is a technical partner — not a server humming in the closet. 

What This Means For Public Affairs

For public affairs organizations, the lesson of Friday is concentration risk. Building your workflow on a single hosted model or relying on vendors who do so, was always risky, but Friday's events made that risk plain for everyone to see. The winners won't be the shops that bet on the right model. They'll be the ones who never bet everything on one. They build for optionality — a primary frontier tool for the hardest work, and a tuned, open-model workhorse that keeps the lights on, whatever happens next.

Two Ways To Operationalize This Email

 🔱 Audit your single points of failure. List every workflow that breaks if one AI vendor goes dark tomorrow. Friday made that a real scenario, not a thought experiment.

🔱 Stand up an open fallback. Find vendors deploying local models as your primary hedge for an uncertain AI future.

🔱 If you're enjoying this content, please consider forwarding this email to a colleague or friend. 

🔱 If you're not already a subscriber, please sign up here to stay up to date on the latest developments in political technology.

🔱 If you have suggestions for future newsletters or any other feedback, we'd love to hear from you.