Spain’s AI law: what happens now that the bill has lapsed
On 6 October 2026 the BOE, Spain’s official gazette, published Royal Decree 806/2026, which dissolves the Congress and the Senate and calls a general election for 29 November. On the same day the Organic Bill on the good use and governance of artificial intelligence lapsed. That was the text meant to become Spain’s AI law. The Congress record now lists it as “Concluded (Lapsed)”.
The practical consequence is simple to explain and somewhat harder to manage. The EU AI Act still binds companies exactly as before, but Spain still lacks the law that was meant to say who supervises each system and how breaches are penalised. In this article I explain what has happened, what still applies, what is left up in the air and what to do in the meantime.
What has happened to Spain’s AI law
The Council of Ministers approved the bill on 26 May 2026, the Government submitted it to Congress on 28 May and it was published in the Official Gazette of the Cortes in June. It spent the whole summer at the amendment stage in the Committee on Economy, Trade and Digital Transformation, with the deadline extended week after week, and it never reached a vote.
With the dissolution, initiatives that have not completed their passage lapse and do not carry over to the next legislature. To revive the law, the next Government will have to approve a new bill and send it to the new Cortes, which will be constituted on 23 December 2026. There is no shortcut before then: with Parliament dissolved no bill can go through, and after the election Law 50/1997 on the Government prohibits the caretaker Government from submitting bills.
What still applies even without a law
The AI Act is an EU regulation and applies directly in Spain. It does not depend on any national law to be binding:
- Since 2 February 2025: the ban on unacceptable AI practices and the obligation of AI literacy for staff. Since 27 July 2026, under the Digital Omnibus, this is an obligation of means: you must take measures, not guarantee a specific level.
- Since 2 August 2025: the obligations for general-purpose AI models.
- Since 2 August 2026: the transparency obligations in Article 50 (generative systems already on the market have until 2 December 2026 to mark their content) and the European Commission’s power to fine providers of general-purpose models directly.
- From 2 December 2027: the obligations for Annex III high-risk systems. The Digital Omnibus on AI, published in the EU Official Journal on 24 July 2026, postponed that date and moved the date for Annex I systems to 2 August 2028. I cover this in the article on the postponement of the EU AI Act.
The use of AI is also still subject to other rules that already have a supervisor and penalties in Spain. The most obvious is the GDPR, enforced by the AEPD, but there are also consumer rules and employment law, which gives workers’ representatives the right to know about algorithms that affect their working conditions.
What is left up in the air
What does not exist is the national piece that the AI Act leaves to each Member State:
- Authorities without legal designation. AESIA has existed since 2023: Royal Decree 729/2023 approved its statute and set its seat in A Coruña. But the full allocation of market surveillance between AESIA and the sector supervisors was in the law that has lapsed.
- A penalty regime still pending. Article 99 of the AI Act requires each Member State to lay down its penalties and notify them to the Commission by the date of application at the latest. In Spain, infringements and penalties require a law passed by parliament, so without that law the ability of Spanish authorities to penalise AI Act breaches is, at the very least, open to question.
- The rules for public administration. The inventory of AI systems used in administrative procedures and the AI officer role that the bill planned for the public sector also lapse.
This should not be mistaken for impunity. The obligations remain in force, the Commission can already act on general-purpose models and the AEPD still imposes penalties whenever personal data are involved, which is almost always. Besides, a breach today will still be a breach once there is a supervisor able to penalise it.
What the bill proposed
The lapsed text is a useful guide to what will probably return, although the next Government may change it. It split supervision by sector:
- AESIA: prohibited practices and most high-risk systems, as well as the single point of contact with the European Commission.
- AEPD: biometric systems and prohibited practices linked to the processing of personal data.
- General Council of the Judiciary: high-risk systems in the administration of justice.
- Banco de España, CNMV and the Directorate-General for Insurance and Pension Funds: creditworthiness assessment, securities markets and pricing of life and health insurance respectively.
On penalties it was aligned with the AI Act: up to 35 million euros or 7% of worldwide turnover for using prohibited practices, up to 15 million or 3% for serious infringements and up to 500,000 euros or 0.5% for minor ones. For banking and insurance the message was clear: the supervisor of their most sensitive AI systems would be the same one that already supervises them as institutions.
What to do now
The lapse of the bill does not change what I recommend to my clients, because almost none of this depends on Spanish law:
- Inventory your AI systems. Which systems are used, for what, with what data and who is accountable. I explain it in the AI asset inventory.
- Check what is already enforceable: prohibited practices, AI literacy and Article 50 transparency.
- Classify systems by risk and identify those that will be high-risk in December 2027.
- Prepare the documentation required by the conformity assessment of high-risk AI systems.
- Set up a management system to support all of the above. ISO 42001 is the natural reference.
Conclusion: without a Spanish law, the AI Act still stands
The lapse of Spain’s AI law delays how supervision is organised, not the obligations. Anyone who uses the delay to shelve compliance will arrive late for December 2027 and with a record that will be hard to justify to whichever supervisor is finally designated. Anyone who uses it to put their AI inventory and governance in order will be ahead when the law returns.
If you want to prepare your organisation for the AI Act with a realistic plan, even while Spain’s own law is delayed, I can help you lead it. Let’s talk.
Frequently asked questions about Spain’s AI law
No. The organic bill lapsed on 6 October 2026 when Parliament was dissolved. The next Government will have to approve a new bill and take it through Parliament from the start.
Yes. It is a directly applicable EU regulation: prohibited practices, AI literacy, general-purpose models and Article 50 transparency already apply, and Annex III high-risk obligations arrive on 2 December 2027.
AESIA has existed since 2023, but in Spain penalties require a law passed by parliament and the penalty regime was in the lapsed bill. Meanwhile the AEPD still enforces the GDPR and the Commission can fine providers of general-purpose models.
Once there is a new Government, which will have to approve another bill and send it to the Parliament elected on 29 November. A caretaker Government cannot submit bills.
Want to see how this applies in your organisation? Write to me.
Book 20 minutes