Bairo GonzalezLeandro · Martinez
Start your guided journey The story

In the world · Artificial intelligence · 29 June 2026

EU AI Act: the timeline and the delay to 2027

The EU AI law has applied in stages since 2024. In 2026 the EU pushed high-risk rules to December 2027. What already applies and what changes.

bairogonzalez.com team, drawing on Bairo's story · Published

The EU AI Act, the European Union's artificial intelligence law, entered into force on August 1, 2024 and applies in stages. In 2026, the EU approved the package known as the Digital Omnibus, which delayed the rules for high-risk systems: from August 2, 2026 to December 2, 2027, and to August 2, 2028 when the AI is embedded in products that are already regulated. The Council gave its final approval on June 29, 2026.

What happened

The European law classifies AI by risk and ties obligations to dates. The original calendar, described by the European Commission, had four steps:

  • February 2, 2025: the prohibitions take effect, such as manipulation that exploits vulnerabilities and social scoring, along with the AI literacy duty for those who use the technology.
  • August 2, 2025: the rules for general-purpose AI models, the large language models, begin. In July 2025, the Commission published a voluntary code of practice to help these companies comply.
  • August 2, 2026: the high-risk and transparency rules were to take effect.
  • 2027: systems embedded in regulated products.

On May 7, 2026, the Council and Parliament reached a political agreement to simplify the law, according to the Council of the EU. Parliament approved it on June 16 and the Council on June 29. The result, according to the analysis by the law firm Freshfields and the coverage by The Parliament Magazine:

  • "stand-alone" high risk (such as AI in hiring, credit and education) moves to December 2, 2027;
  • high risk embedded in products moves to August 2, 2028;
  • the transparency rules stay at August 2, 2026, with a deadline of December 2, 2026 to label AI-generated content in systems already on the market;
  • a new prohibition is added: AI that generates child sexual abuse material or non-consensual nudity.

Why it matters

The delay is not a retreat. The prohibitions and the rules for large models already apply, and so does transparency. What gained time was the heaviest part: documentation, risk management and human oversight of high-risk systems.

For Brazil, the EU AI Act matters in two ways. The first is trade: whoever sells an AI-powered service to customers in Europe answers to European law, wherever they are. The second is influence: Bill 2338/2023, passed by the Senate, follows the same logic of classifying by risk. Preparing for one is, to a large extent, preparing for the other.

In Bairo's view

In the reading of Bairo Leandro Gonzalez Martinez, Europe put into law what he defends as a principle: AI is an extension of the human, and that is why it needs a responsible human where the decision carries weight. For him, the delay shows that even regulators learn on the move, and that those who build should not wait for the deadline to do the right thing.

Bairo bets on niche AIs for this reason too. An agent that performs a bounded task has a risk that is easier to measure, document and audit than a system that tries to do everything. CLAIN, the agent orchestrator he founded, is conceived this way: small specialists, an orchestrator that brings the result together and a person who decides. The goal of 142 niche AIs in 142 days is a goal, and each new cell is a chance to be born with the European yardstick already in mind.

Where this meets the ecosystem

Several of Bairo's ventures have a foot in Europe and the United Kingdom, which is why the European calendar enters the planning:

  • Xperienc Global Labs, registered in London, proposes teleconsultation with a clinical assistant and a second opinion, not a diagnostic report, validated by a professional. Health is a high-risk area under the European law.
  • CEASA Bank, a financial platform for agribusiness that operates on the infrastructure of an authorized institution, proposes pricing and export for buyers abroad, including Europeans.
  • qrqbits works on file protection with post-quantum cryptography, a topic that speaks to the European duty of system security.

None of them claims compliance with the EU AI Act. What the European law offers today is a map, and the proposal is to start walking it now.

Sources

Read also

More on this topic

In the world: Artificial intelligence