Fable 5 was pulled but should such powerful models be available in the first place? 

Within days of release, Anthropic disabled access to Fable 5 and Mythos 5 after the US government ordered it to suspend access by any foreign national. Anthropic said the practical result was pulling both models for compliance.  

The reported concern was a potential way of bypassing the model’s safeguards. Anthropic has pushed back, saying the government did not give detailed evidence, that the issue appeared narrow rather than universal, and that similar vulnerability-finding capabilities are already available in other public models, including OpenAI’s GPT-5.5, which remains available.  

But the bigger story is not about one model’s vulnerability. It’s about what this moment reveals about AI sovereignty, governance, and whether such powerful tools should be public at all. 

The sovereignty issue gets real 

For years, the debate about AI sovereignty has often sounded more technical and less like something for the everyday user to think about. It’s been about compute, chips, cloud dependency, data centres, foundation models, and regulatory influence. All of that still matters but the Fable 5 episode crystallises the sovereignty dilemma. 

If a model can be switched off globally because one government decides access should be restricted, then every other country, business, and public institution has to ask: how much of our operating capacity depends on someone else’s political decision?  

The uncomfortable answer is that many organisations may not know until access is gone. That’s the real risk Fable 5 exposes. As AI moves from experimentation into core workflows, dependency can build quietly, and quickly. Sovereignty, then, is not just about where models are built, but whether countries and businesses have the governance, skills, and contingency plans to avoid being caught out by decisions they don’t control. 

Businesses shouldn’t chase power for its own sake 

There’s also a practical lesson here for organisations racing to adopt the latest model. 

“We’re in an arms race over model functionality and release amongst the tech companies,” says David Pool, Data & AI Development Director. “Most enterprises don’t need the functionality of the latest models. You don’t need a Formula 1 car to do your weekly supermarket shop,” he explains. 

That point matters because much of the current AI conversation assumes that more powerful always means more useful. For many organisations, the smarter question is not, “How do we get access to the most advanced model?”, but “What capability do we actually need, how should it be governed, and what happens if access changes?” 

Businesses can maintain some control if they understand which tasks genuinely require frontier models, which can be handled by smaller or more controlled systems, where sensitive data sits, what human oversight is required, and how workflows continue if a provider withdraws, restricts, or changes access.  

Powerful models need rules 

As models become more powerful, should every capability be deployed to the general public at all? 

The temptation is to frame this as a simple fight between freedom and control. The reality is that powerful AI needs oversight. 

Vicky Crockett, Portfolio Director, Artificial Intelligence, likens AI regulation to controlling nuclear weapons. Both, she argues, have the capacity to cause mass destruction. But with nuclear weapons there are clear rules about who controls uranium resources. Those rules are far less clear when it comes to AI. 

“If everyone has access to smaller, faster inference models, who’s to stop anyone, be it ordinary citizens or terrorist groups, from using them irresponsibly? It’s vital that governments agree internationally about who has access to these models.” she says. 

Vicky also argues that regulation will make AI use more attractive, not less. “If there aren’t safety systems around it, people will reject it or misuse it in a way that’s an overall disadvantage to humanity.” 

Transparent, robust regulation can be the first step to clarity 

The Fable 5 episode shows why regulation shouldn’t be seen as the enemy of innovation, but as the scaffolding that helps it scale safely. Clear, internationally aligned governance would give users, businesses, and governments confidence that the tools they rely on are safe, valid, compliant, and not likely to disappear through opaque decision-making.  

That reassurance matters because people are more likely to adopt AI when they understand the rules and trust the safeguards.