Man, the hits just keep on coming. Anthropic has now published a statement saying the US government has directed it to suspend access to Fable 5 and Mythos 5. Not just outside the US, I hasten to add. According to Anthropic, the directive applies to any foreign national, whether inside or outside the US, including foreign-national Anthropic employees. Anthropic says the net effect is that it must disable Fable 5 and Mythos 5 for all customers in order to comply. That is not a small product-access issue anymore. That is the state treating frontier Artificial Intelligence as a national-security asset. And this is where the downstream consequences become interesting. Anthropic has spent years arguing that frontier models require serious safeguards, serious oversight, pre-deployment testing, red-teaming, risk thresholds, and the ability for governments to block unsafe deployments. Now the government has intervened, and Anthropic is arguing that the action is opaque, disproportionate, and not grounded in the technical facts. Be careful what you wish for. The governance question here is not whether powerful AI systems need oversight. They do. The question is whether that oversight is transparent, technically evidenced, contestable, and proportionate. Because once frontier AI is framed as a national-security object, the centre of gravity completely changes. It moves from product governance to state control. From safety documentation to licensing regimes. From responsible deployment to geopolitical access management. And from “trust us, we have safeguards” to “prove the safeguards, prove the risk, prove the proportionality, and prove the decision process.” That is the real lesson here. Governance cannot just be asserted when it is useful and challenged when it becomes inconvenient. There is also a strategic lesson here for those of us in the UK and the European Union. If access to frontier models can be disrupted overnight by US export-control action, then the UK and the EU should treat this as a serious wake-up call. We need sovereign AI capability across the full stack: energy, compute, research labs, infrastructure, talent, model development, and the application layer built on top of that. The talent is here. The knowledge is here. The market need is here. What is required now is the political will, industrial strategy, and more, much more capital allocation to build at scale. And there is an added irony. Anthropic’s own statement says Fable 5 was red-teamed for thousands of hours with the United Kingdom Artificial Intelligence Safety Institute (UK AISI), the US government, third parties, and internal teams. Yet despite that, access can still be switched off through a unilateral US directive. That should concentrate minds in London and Brussels. Strategic dependence always looks manageable until the day it is exercised.