Explore. Build. Grow
Explore. Build. Grow

Our flexible three-stage framework to craft digital experiences.

Learn more
Expertise
Outcomes
View our portfolio
View our portfolio

Explore our range of digital solutions.

Learn more
Industries
Our Approach to AI
Our Approach to AI

Keeping ourselves at the forefront of emerging technology is part of our DNA. But we recognise that ‘technology for technology’s sake’ is not the end goal – there is a balance to be struck.

Learn more
DXP
Headless
Cloud Platforms
Marketing Optimisation
Contact Us
Contact Us
Explore. Build. Grow
Explore. Build. Grow

Our flexible three-stage framework to craft digital experiences.

Learn more
Expertise
Outcomes
View our portfolio
View our portfolio

Explore our range of digital solutions.

Learn more
Industries
Our Approach to AI
Our Approach to AI

Keeping ourselves at the forefront of emerging technology is part of our DNA. But we recognise that ‘technology for technology’s sake’ is not the end goal – there is a balance to be struck.

Learn more
DXP
Headless
Cloud Platforms
Marketing Optimisation
Contact Us
Puppeteer hand
Home / Insights / ＿

Washington switched off the frontier. Did your business even notice?

A blackout of Anthropic’s most powerful AI models is a free fire drill for anyone betting their business on a single model.

Adam Griffith
Adam Griffith

14 June 2026

5 minute read


On Saturday morning I was at MiniRoos, watching my son referee, with half an eye on my laptop back home via Claude’s mobile app and its Remote Control feature. We're building an internal tool at the moment, and I was using Fable, Anthropic's newest model, to push through a feature between whistles.

Then the build stopped dead. 'Model unavailable.' So I did what everyone does with an error they don't understand: I pasted it into Google, which sent me straight to Reddit, where half the internet was already going bananas. The model hadn't been throttled or rate-limited. It was gone.

I couldn't just duck home and fix it. You can't switch models from the phone app (Anthropic, if you're reading: feature request!), and I was needed on the sideline, because a junior ref occasionally needs protection from the more competitive parents… So I rang my wife and talked her through changing the model on my laptop, mid-match, so the work could limp on.

Funny, in hindsight. A lot less funny once I read why it had happened.

Three days. That's how long Anthropic's most powerful model lasted in the open.

“
Your most important tool can be switched off by a government you didn’t elect, for reasons you’re not told, with no notice and no appeal.

On 9 June the company released Fable 5, the most capable model it has ever put in public hands. Alongside it came Mythos 5: the same model with its safety limits lifted, handed to a small group of vetted cyber-defenders. On 12 June the United States government ordered Anthropic to cut every 'foreign national' on the planet off from both models, inside American borders or out, its own non-American staff included. You can't enforce a line like that by halves. So Anthropic switched both models off, for everyone, everywhere.

If that sounds like the opening of a thriller, the reality on the ground was duller, which is rather the point. For all the headlines, most businesses in this country didn’t feel a thing. Every other model kept running. The marketing drafts still got drafted, the support bots still answered, the analysts still got their summaries.

Did the blackout matter on the day? For most of us, no. Does it matter for how we build from here? Far more than the outage itself.

A foreign off switch

Start with how it happened, because the mechanism is the warning. The lever Washington pulled is an old one called ‘deemed export’: showing controlled technology to a foreign national, even one sitting at a desk in California, counts in law as exporting it. Point that rule at an AI model and you can switch it off for the entire world without it ever leaving the country.

Being a trusted friend bought nothing. Australia sits in the top tier of Washington’s AI chip rules, the group of allies that faces no restrictions. The model directive ignored the tiers completely. It didn’t care where you were or how friendly your government was. It cared that you weren’t American.

Here is the part worth sitting with, wherever you are. Your most important tool can be switched off by a government you didn’t elect, for reasons you’re not told, with no notice and no appeal. That isn’t an innovation problem or a procurement problem. It’s a continuity problem, and it has just stopped being hypothetical.

Now the fair counter-argument. Isn’t this temporary? Probably. Anthropic calls it a misunderstanding and says it is working to restore access, and by the time you read this the models may well be back. Good. But a fire drill doesn’t have to burn the building down to be worth your attention. The precedent is the lesson, not the downtime.

The habit the blackout exposed

So why did so many businesses sail through? Mostly because they were never running on the frontier in the first place, whether they realised it or not.

The ones who got caught were the people who live closest to this stuff (ah, like me 🙋‍♂️). Get close enough and you pick up a reflex: reach for the biggest, newest, most expensive model for everything, on the assumption that more capability is always better. There's even a word for it now, 'tokenmaxxing'. It feels responsible. It's usually a waste.

Most of what businesses actually ask AI to do, summarise a document, draft an email, sort enquiries, answer a customer, classify a form, does not need a frontier model. A smaller, cheaper, older one does the job just as well, often faster, at a fraction of the cost. The frontier earns its keep on a narrow band of genuinely hard problems. For the rest of the work, paying for it is like hiring a barrister to write your out-of-office reply.

The businesses that didn't notice the blackout were, by and large, the ones already matching the model to the task. They got resilience as a side effect of spending sensibly. That isn't luck you want to rely on. It's a discipline worth choosing on purpose.

What to actually do about it

None of this calls for panic, and it certainly doesn’t call for swearing off American AI, which would be both impractical and daft. It calls for the kind of boring resilience any business applies to a critical supplier it doesn’t control.

Build so you can swap. An abstraction layer between your product and the model means changing provider is a configuration change, not a rebuild. Keep a fallback you’ve actually tested, not one that lives on a slide, because you don’t want to be meeting your plan B for the first time at 5.21pm on a Friday (or Saturday morning at MiniRoos!). Right-size relentlessly: route each job to the cheapest model that does it well, and let the harder jobs go to the bigger ones. Done properly, running several models is how you get better results and lower bills, not just insurance.

And widen the bench. That includes the strong open-weight models now coming out of China, which are cheap, increasingly capable, and good enough for plenty of everyday work, provided you go in clear-eyed about where your data goes and what you’d actually trust them with. You don’t have to use them. You do want to be able to.

The blackout itself will be a footnote within a month. The dependency it revealed won’t be. The question it leaves on the table has nothing to do with Anthropic: when someone next decides your AI should go dark, in Washington or anywhere else, does your business keep working?

Best to have an answer ready before you need it.

Image: Sivani Bandaru, Unsplash

Adam Griffith
Adam Griffith

14 June 2026

5 minute read

Enjoyed this article? Share it...

Want to know more about Luminary’s AI Readiness Framework?

Get in touch

Keep Reading

Want more? Here are some other blog posts you might be interested in.

Follow
Contact Us
1300 554 891
hi@luminary.com

Our locations

Stay informed
Our best articles, observations and news delivered to your inbox
Subscribe
Flags

We acknowledge the Traditional Custodians of the land on which we work, live and engage. We pay respect to Elders past and present.

Luminary is a registered trademark of Luminary Digital Pty Ltd

Terms of Services
Privacy Policy