Why Claude Fable should have us thinking about AI Sovereignty

Claude Fable was here, and now it’s gone. At least as of this writing.

The release of Claude’s Fable, a supposedly consumer-ready version of their supposedly super dangerous Mythos model, came as a surprise when it popped up in my options. I had to wonder, if after a few weeks and warnings of Mythos’s power to break cybersecurity systems, how it was so quickly turned into something I could have easy access to.

Apparently I wasn’t the only one wondering. Amazon CEO Andy Jassy reported to the White House that they had found a way to jailbreak Fable. This resulted in the White House issuing an order that Anthropic, the AI company behind Claude, had to make Fable inaccessible to any “foreign nationals” for security reasons. Because it would be difficult to track each individual account and user and their nationality, Anthropic removed access from Fable for everyone.

Article continues after this advertisement

That’s the backstory. But the question is, why does this matter and what are things that we should be considering especially given our context?

FEATURED STORIES

TECHNOLOGY

TECHNOLOGY

TECHNOLOGY

So what if Fable disappears?

As a consumer, there isn’t really that much impact. I got to use it the few days it was available, but it wasn’t around long enough for me to build my workflow around it. There wasn’t any reliance. I’m sure most Claude users felt the same: not enough time with the model to fully test, maximize, and integrate into workflow.

RELATED STORY: OPINION: AI legislation taking a wrong turn in the Philippines

For the everyday Claude consumer, there won’t be that much impact really. We had access to a really powerful model, and now we don’t. Back to Opus for us. For other chatbot users, this doesn’t really mean anything.

Article continues after this advertisement

But I do believe there are bigger signals that this sends.

AI Access and Policy Instability

In early 2025 the USA’s vice-president, JD Vance, gave a speech that was aggressively anti-AI regulation, claiming that such regulations would limit AI development. That was an initial declaration of AI policy, and a sharp turn from their previous administration’s more…let’s say thoughtful approach to AI. It has also been a theme of the Trump administration to speak of AI in terms of not just global leadership, but dominance. I know they might use a lot of words and may not necessarily mean them, but the consistent use of domination in their language around AI should raise eyebrows at least.

RELATED STORY: US Government orders Anthropic to shut down Fable 5 and Mythos 5 globally

Article continues after this advertisement

Take that as a first ingredient. Now add the whole narrative around AI development as an arms race between the US and China. As a last piece, consider the recent steps the Trump administration is making to now regulate and control AI. From “hands-off” to the recent Executive Order asking AI companies to submit their models for review to where we are now with the Fable and Mythos ban due to “national security” risks.

What I’m trying to establish here isn’t a clarity of AI Policy, but rather an instability in said policy. If you’re only thinking about this as an individual consumer, you wouldn’t have much to worry about. But if you are thinking of this as someone managing AI-integrated workflows, running a company that uses AI, and most especially if you are concerned with AI Policy or seeing AI integrated into government processes, then this should be a real cause for concern.

As one of the leading thinkers in AI, Gary Marcus, says, “Thue (sic) rushed and seemingly unprincipled decision will leave the rest of the world rushing towards “sovereign AI” (made for example by European or Canadian companies) and even Chinese AI, rather than deal with the uncertainty in the US, in which any model can be shut down at any time without warning and without clear explanation.

Simply stated, if the US government can block access to a specific model without needing much evidence or any specific policy, then what does this say about the vulnerability of any AI, or even other software products for that matter? Perhaps previously it was a safe assumption that our access to AI and other tools were just business transactions and weren’t being touched by politics. But not anymore.

We can see that even before this other countries have been trying to remove their reliance on US tech stacks. For example France is shifting from Windows to Linux. South Korea, thanks to its already powerful tech industry, is building out its own AI infrastructure and models.

This then isn’t just about the threat of the United States blocking access. Nor is it about turning around and embracing Chinese AI, especially given our current tensions and the vulnerabilities of adopting their tech might create for us. It’s advancing the idea that we need to craft our own approach to Sovereign AI.

Sovereign AI enters the chat

The core idea is that whatever AI exists now wasn’t built by us, doesn’t reflect our culture or values, and is kind of an imposition. Further, if we want to have sovereignty, our systems can’t be beholden to other nations. What happens if there is geopolitical conflict and suddenly, as in the case of Fable/Mythos another nation cuts off its services? We need to have our own infrastructure and systems that we have control over.

Often when I speak to people about Sovereign AI, either the conversation gets stuck around how much more powerful and far ahead other countries are, or our own inability to compete at building frontier models given the cost, infrastructure, and staffing that would be needed. I think it’s fair to point out our limitations. But these assumptions are often based on how we have seen the US and China build their AI. As part of Sovereign AI, we need to have our own strategy and our own approach that might draw from what we have seen from other countries, but is ours alone based on our resources, capacities, and identified opportunities.

I’ll outline some approaches that I think we should consider. I recently got to attend an AI Policy and Governance training built around South Korean approaches, and one of the things they kept emphasizing was that it was most important to believe that we could change. I know that might sound naive, but starting from that core empowering belief, we can start to imagine how we might want to see AI develop for us as Filipinos.

First off, I think we should be looking at Green AI or Sustainable AI as core to any kind of AI that we build. If we take the hyperscaler approach that the US is taking, we would be trashing our environment. Next, the government needs to focus not on building frontier models, but on deploying small models developed for specific uses. The assumption here is that we can build out these models that can inspire startups to build new businesses. Next is that we create AI through consultation with communities where it will be deployed, so that we know what those communities need and understand the kind of systems that would help them best.

Your subscription could not be saved. Please try again.

Your subscription has been successful.

This approach is counter to the “do-everything” machines we currently get from the major chatbot providers. But again, given this new context where it’s possible that those systems could get pulled out from under us, this is the exact moment when we should start considering new approaches and new ideas, like Sovereign AI. Do note that what I’ve offered here is only a cursory introduction to the ideas, but it’s also an invitation for people to read further about what this might mean and maybe even advocate for it more.