What comes into effect on August 2, 2026
The AI Act (EU Regulation 2024/1689) follows a phased implementation since its entry into force in August 2024. On August 2, 2026, the obligations of Article 50 — the transparency requirements — become enforceable for all companies operating in the EU or whose AI systems affect European users, regardless of where the company is headquartered.
In practice, this means three concrete obligations: first, any AI system that interacts with users — chatbots, virtual assistants, automated customer service systems — must clearly inform the user that they are interacting with an AI, not a human. Second, AI-generated content that is published to inform about matters of public interest must be clearly labeled as such. Third, video and audio deepfakes — manipulations of real people's images or voices — must be identified as synthetic.
The fines for non-compliance are substantial: up to €35 million or 7% of the annual global turnover for violations of prohibited systems, up to €15 million or 3% of turnover for violations of other obligations.
What was postponed — and why
The European Parliament voted in April 2026 to postpone the compliance deadlines for high-risk systems: from August 2026 to December 2027, with specific sectoral obligations being delayed to August 2028. The legal vehicle was the "Digital Omnibus" — a regulatory package that revised multiple digital directives simultaneously.
High-risk systems under the AI Act include AI used in critical infrastructure, biometrics, employment and recruitment, education, access to essential services (credit, housing), law enforcement, and administration of justice. These are exactly the AI uses with the greatest potential for negative social impact — and where regulation matters most.
The pressure for postponement came from two simultaneous fronts. Big techs — especially Microsoft, Google, and Meta — argued that the technical requirements for high-risk compliance were impractical within the original timeframes and threatened European competitiveness against the US and China. The Trump administration, in turn, signaled that a strict implementation of the AI Act would be interpreted as a non-tariff barrier to American trade, with potential implications for broader trade negotiations.
The result was a political agreement in the EU Council in May 2026 that maintained the transparency deadlines but eased the high-risk ones. The European Commission described the move as a "pragmatic calibration", not a retreat. Critics call it capitulation.
What this means for Brazilian companies
The AI Act has an extraterritorial reach: any company that places AI systems on the European market or whose AI outputs affect people within the EU is subject to the regulation, regardless of where it is headquartered. Brazilian companies with operations or clients in Europe — especially in the financial, health, and human resources sectors — need to map their AI systems in light of this regulation.
The so-called "Brussels effect" — the tendency for European standards to become global standards due to the EU's commercial weight — suggests that the AI Act will shape AI practices globally even where it is not formally applicable. Companies that develop for the European market incorporate the requirements into their global products; suppliers that serve these companies adopt the same practices due to value chain pressure.
The complete implementation calendar
February 2025 — prohibitions of unacceptable risk systems (like social scoring and subliminal manipulation) came into effect. August 2025 — governance requirements for general-purpose AI (GPAI) models with systemic risk, including documentation, risk assessment, and serious incidents reporting. August 2026 — transparency obligations of Article 50 (chatbots, AI content labeling, deepfakes). December 2027 — high-risk systems (postponed from the original deadline of August 2026). August 2028 — specific high-risk sectoral obligations.
For companies using AI in regulated products (medical devices, vehicles, toys), compliance with the AI Act integrates with the existing CE certification process — an additional layer of documentation and evaluation, not a separate process.
