As of August 2, 2025, the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) enters a pivotal phase with the implementation of several key provisions aimed at strengthening oversight of high-impact AI systems. This milestone marks the start of formal enforcement for prohibited AI systems and the regulation of general-purpose AI models, alongside the possibility of sanctions for serious non-compliance.
One of the most significant developments is the activation of the enforcement regime under Chapter XII of the regulation. National and EU authorities are now empowered to impose penalties for violations such as the use of prohibited AI systems, as defined in Article 5. These include technologies that subliminally manipulate human behavior, systems for social scoring, and real-time remote biometric identification in public spaces.
Although the European Commission’s direct sanctioning powers over providers of general-purpose AI models will not apply until August 2026, enforcement actions for other serious breaches can already be initiated.
Also entering into force is Chapter V, which sets out obligations for general-purpose AI models, including large language models. Providers of these models must now comply with requirements related to technical documentation, transparency, and safety and risk mitigation, particularly when their models pose systemic risks.
In parallel, on July 10, the European Commission received the first Codes of Practice for general-purpose AI. These voluntary frameworks support supervised self-regulation and aim to ease compliance with the AI Act. Structured around copyright, transparency, and safety and protection, the codes have already been adopted by major industry players such as OpenAI.
What’s Next?
The rollout of the AI Act will continue in stages over the coming months:
- Throughout 2025 and 2026, additional chapters will come into effect, particularly those governing high-risk AI systems, with more detailed technical requirements and mandatory conformity assessments.
- From August 2026, the European Commission will gain direct enforcement powers over providers of general-purpose AI models, including the ability to impose fines.
- EU Member States, including Spain, must adopt complementary national legislation to designate competent authorities, establish oversight procedures, and define complaint mechanisms.
- The role of the European AI Office will be further consolidated, becoming central to cross-border coordination, technical supervision, and the development of harmonised standards.
This phased implementation not only enhances legal certainty in the AI landscape but also presents new challenges for developers, providers, and regulators operating in an increasingly structured and demanding environment.