Brussels/Rome, October 2025 – 2025 is proving to be a pivotal year for Artificial Intelligence (AI) regulation globally, with the European Union consolidating its regulatory leadership through the progressive implementation of its pioneering AI Act. This regulation, which entered into force in August 2024, aims to establish a legal framework that ensures AI systems are safe, ethical, and respectful of fundamental rights, impacting businesses and governments worldwide.
Regulatory Integration: GPAI and Unacceptable Risk
The European approach, based on risk classification, is now the focus of international attention. The summer and autumn of 2025 marked crucial milestones in the implementation of the law.
First, obligations for providers of General Purpose AI (GPAI) models, such as advanced language models, came into force on August 2, 2025. Providers of models deemed to pose a "systemic risk" must now implement rigorous measures, including security testing (or adversarial testing), risk mitigation, and reporting serious incidents to the European AI Office.
Furthermore, the regulation explicitly establishes what is prohibited. AI systems that pose unacceptable risks, such as those using social scoring or manipulative techniques to exploit the vulnerabilities of groups of people, are explicitly prohibited from February 2025. Fines for violating these bans can reach up to €35 million or 7% of annual global turnover, whichever is greater.
High-Risk Systems and New Obligations for Businesses
The category with the greatest impact on businesses is High Risk. This includes AI systems used in sensitive sectors such as healthcare, education, employment, justice, and the management of critical infrastructure.
For these systems, the law imposes onerous obligations: a conformity assessment is required before being placed on the market, they must ensure human oversight, traceability, and the adoption of rigorous data governance practices to prevent bias. Companies must also register in an EU database. Failure to comply with these requirements can result in fines of up to €15 million or 3% of global turnover.
For Limited Risk systems (such as chatbots or AI-generated content), however, the primary obligation is simply transparency: users must be informed that they are interacting with an artificial system.
The Brussels Effect and the Global Response
The AI Act, which comes into full force in 2026, is establishing itself as the world's first comprehensive legal framework for AI, generating what is being called the "Brussels Effect." This approach is pushing other jurisdictions to reconsider their standards.
Even at the national level, the impact is immediate: Italy, for example, approved its first law on Artificial Intelligence (D.D.L. No. 1146) in September 2025, considered an urgent step to provide a unified vision before the full implementation of the EU Regulation.
Meanwhile, in Asia, countries like South Korea are responding with "sovereign and independent" AI initiatives, confirming that technology governance is now a central issue in geopolitics and global trade. The AI Act is not just a law, but a model for technological governance that seeks to balance the potential of innovation with the protection of fundamental values and rights.
(Associated Medias) - All rights reserved