Pedro Sánchez has a message for the companies building the world’s most powerful AI: don’t expect to mark your own homework. Speaking at an AI regulation event in Madrid on Monday, September 21, Spain’s prime minister said the technology cannot be self-regulated by the people who control it, and laid out a 12-month plan to back that up. The Spain AI regulation roadmap combines tougher oversight with a proposed AI “gigafactory” and new models built with the Barcelona Supercomputing Centre. A gigafactory, for anyone new to the term, is a giant computing site filled with specialised chips, designed for training the most demanding AI systems.
A case against letting AI companies police themselves
Sánchez’s reasoning was simple. Voluntary self-policing has failed before, he argued, and the people running the leading AI firms have incentives that will never fully match those of the public. So he wants strict rules, plus a national pact, which he called a “gran acuerdo de país”, that would bring business, workers, unions, political parties and public bodies to one table. Invitations are due to go out in the coming weeks.
He also promised a firmer cybersecurity posture. The government will reinforce its defences against frontier AI, the newest and most capable systems. Reuters’ English report says he described those models as “aggressive”, though its Spanish version uses a word closer to “offensive”. Specific measures weren’t spelled out. AI and security are already colliding elsewhere, as HCN’s earlier report on Google, Gemini AI and the hacked companies showed.
Data centres got their own attention. Sánchez wants them held to high energy and environmental standards, built so Spain keeps control over the data inside, and useful to the communities around them. Reuters’ Spanish report adds a line about not exhausting scarce resources. The plan also promises new models with the Barcelona Supercomputing Centre aimed at climate, health and energy problems.
The gigafactory has been in the works for a while
This part isn’t new. Brussels wants three to five giant AI computing sites across the EU, each with around 100,000 advanced chips, and has set aside €20 billion to help fund them. Spain has been in the running for more than a year with a two-location bid: Móra la Nova in Tarragona and San Fernando de Henares near Madrid. The estimated cost is about €5 billion, with roughly 70 percent expected from private investors. In June, the cabinet approved €719 million for it, and Telefónica, ACS and Banco Santander are among the companies involved.
Winning isn’t guaranteed. In its July statement, the government said it was still waiting for the European Commission to open the official call for proposals. Monday’s coverage doesn’t state that the gigafactory in the Spain AI regulation roadmap is this same bid, but it’s the obvious candidate. Whether Brussels chooses Spain hasn’t been decided.
The industry hasn’t answered yet
In the coverage available at the time of writing, there was no response from AI companies, EU officials or outside researchers.The timing adds context. Reuters noted that Sánchez spoke just days after the heads of major US AI labs called for a slowdown, warning that AI could soon start improving itself and get beyond human control. Those calls are voluntary and conditional. AP reports that Anthropic has said it would slow or pause only if its rivals did the same, in a way that could be verified. MIT Technology Review was more doubtful, saying nobody is clear on what a slowdown would look like in practice and that the companies have their own reasons to appear responsible while investors watch.
Sánchez’s line runs the other way: good intentions from the companies aren’t enough.
Why the Spain AI regulation roadmap matters
Spain is not starting with a blank page. It already has a national agency for supervising AI, and EU-wide AI rules apply on top of that. So what the speech offers is direction, not a new rulebook. A large EU government is arguing publicly that stricter oversight and home-grown computing power should go together. One thing to keep in mind: the idea that a gigafactory would let Europe train and run AI without relying entirely on foreign providers is the Spanish government’s own argument. It hasn’t been tested. The speech and the main points of the roadmap, however, are well documented by Reuters and Spanish media.
What’s still missing from Spain’s AI regulation roadmap
There’s no draft law yet, no total budget for the plan and no named body to enforce it. Strict rules are a promise, not a legal text. The environmental language around data centres also sits uneasily with the fact that they use a lot of electricity, and the reports don’t explain how the standards would be enforced.
[…] its core, the Airbus PARACOM contract is about controlling how information flows across a military network that is not really one network at all. France’s armed forces, like […]