AI is not a product, it is a weapon. And Europe has none

A few days ago the Government of Spain showed its chest announcing that the EU AI Regulatory Sandbox project had been a success. It is difficult to understand why, but while these efforts to regulate require notable investments (in this particular project, 4.3 million euros), there is a stark and forceful reality:

Spain and the EU are light years ahead of the US and China in terms of innovation in AI. And in this situation, how will the EU survive?

Chronology of events. That was already a problem before, but each new advance of the US or Chinese frontier models makes that hole seem bigger and bigger. We have seen it with what happened to Claude Fable 5which has been vetoed following the order of the US Government. The recent chronology allows us to explain what happened:

• April 8, 2026: Anthropic releases Claude Mythos Previewa model that according to them is so powerful in terms of cybersecurity that they avoid releasing it publicly. Your capacity is being confirmed by companies like Mozilla, which is one of the privileged to be able to use it.
• June 10, 2026: Anthropic launches Claude Fable 5a kind of layered version of Claude Mythos Preview. According to the benchmarks, the model is a qualitative leap, but it has several security measures to avoid being used maliciously. Its availability was initially public in the different plans (free and paid) of Claude.ai. As of June 22, it would only be possible to access it through the API. That was the plan, at least.
• June 12, 2026: The US Government forces Anthropic to turn off the tap and not allow anyone who is not a US citizen to access Fable 5. Anthropic then decides cut off access to absolutely everyone to this model.

AI is already a weapon. All this development of events, still in full effervescencedemonstrates something important: AI has stopped being just a tool and has become a technological weapon. It is the confirmation of that Oppenheimer moment which we already alluded to in the past.

Oppenheimer Moment. What happened with the atomic bomb – a technology that the US Government ended up appropriating – is partly happening now with models like Fable 5: it is the state that ends up having the power, whether we like it or not, so that the world can use it. Its creators, Anthropic, have practically no say even though they have already tried to defend their position. with the scandal of the Pentagon and the Department of Defense.

I’ll see who I let use the AI. The US decision to cut off access to a commercial AI model is unprecedented. Hiding behind the protection of National Security, the Government of this country has decided that Fable 5 was simply too dangerous for other countries to use. He has exerted absolute control over a disturbing technology, and he has done it without anyone being able to do anything. That access to AI can be vetoed overnight due to government imposition once again makes it clear that those who invest in these models have a strategic advantage that can be definitive.

China takes another stance. In fact, China is currently proposing a radically opposite strategy: open weight AI models are the norm among its large companies and its startups. DeepSeek, Qwen, Kimi or Xiaomi Mimo are some of the clear examples of how it is possible to develop advanced AI models with another approach. At the moment China has not vetoed any of these models, and all of them have become a great resource for companies around the world to improve and use. Even North American companies do it: Cursor’s Composer 2.5 model It is nothing more than a derived version of Kimi 2.5.

AI as a commercial weapon. What we are seeing is another episode of the great trade war that both the US and China are maintaining at a technological level. Both play with their resources to put pressure on the other party—or the entire world. But before, the pressure was exerted with things like advanced AI chips or rare earths. Now the US has also decided that the AI ​​software itself (the models) can be used in the same way. And the companies that develop them don’t seem to be able to do much about it.

Spain boasts about what it shouldn’t. But as we said, in Spain what we have are announcements of regulatory efforts. It is true that the intention is surely good and that the idea is reasonable (to develop the technology responsibly), but that focus has proven to be a problem: while Spain boasts of being a pioneer in the application of AI regulation in Europe, our country barely has its own tools in this area. ALIAthe model focused on co-official languages, remains in a surprising background, and what is We have an agency, AESIAwhich once again demonstrates that ambition to regulate while neglecting the other part: that of innovating.

Europe is not much better. What is happening in Spain is almost a reflection of what we are seeing in Europe. Only the French company Mistral is trying to advance with its own AI models, but for now its developments They are very far from the frontier models of the US or China. In Spain we have Magnificbut this startup is not so much a developer of its own models as an aggregator of those already existing in the creative field.

Atomic bombs on one hand, forks on the other. It is probably an exaggeration to compare Fable 5 with the atomic bomb, but one thing is clear: the US has AI models that are effectively far above any European development. It is in a sense as if they had a bomb and Europe was trying to compete with a fork. What experts have been saying for almost three years is that the path chosen by the EU is a mistake. What is needed is investment in innovation, and not with government projects, but with commitments to private companies that imitate the strategy and cycles of the American or Chinese companies. Less bureaucracy, more facilities.

But for now, the EU is not listening. At most, he complains.

In Xataka | The world’s leading expert in identifying deepfakes has a big problem. No longer able to identify them