EU sticks to timetable for the AI Act

The European Union is steadfastly adhering to its timeline for the introduction of the AI Act. Despite intense lobbying from over a hundred technology companies, including industry giants like Alphabet, Meta, Mistral AI, and ASML, who feared a delay and cited adverse effects on Europe’s competitive position, the European Commission has made it unequivocally clear that there will be no interruption, postponement, or grace period. According to a report from Reuters, European Commission spokesperson Thomas Regnier emphasized that the legal deadlines are set in a legislative text and will be respected.

The AI Act is a risk-based regulation aimed at making the development and use of AI systems safer, more transparent, traceable, non-discriminatory, and sustainable. It divides AI applications into different risk levels, imposing varying requirements accordingly. The ban on AI applications with unacceptable risk includes practices such as cognitive behavioral manipulation of people or vulnerable groups, for example, voice-activated toys that could encourage dangerous behavior in children. Also banned is so-called social scoring, which is the classification of individuals based on their behavior, socioeconomic status, or personal characteristics. Systems for biometric identification and categorization of individuals, as well as real-time facial recognition in public spaces, are also prohibited, although there may be limited exceptions for law enforcement in serious cases. These prohibitions have been in effect since February 2, 2025.

Another category consists of high-risk AI systems that could have a negative impact on safety or fundamental rights. This includes AI in products such as toys, vehicles, or medical devices. AI used in critical infrastructures, education and vocational training, employment, access to essential public and private services, law enforcement, migration and border management, or the administration of justice also falls into this category. These high-risk systems must be assessed before being placed on the market and throughout their lifecycle, and must be registered in an EU database beforehand. The obligations for these systems will become fully applicable 36 months after the law enters into force, i.e., by August 2026. For AI applications with limited risk, such as chatbots and generative AI models like ChatGPT, lighter transparency obligations apply. These include disclosing that the content was generated by an AI, designing the model to prevent the generation of illegal content, and publishing summaries of copyrighted data used for training. AI-generated or modified content, such as deepfakes, must also be clearly labeled as such.

The law also promotes AI innovation and startups in Europe by allowing companies to test general-purpose AI models in a simulated real-world environment before they are released to the public. The EU passed the AI Act in June 2024. The full regulations of the AI Act will be fully applicable by mid-2026. The European Parliament has established a working group to monitor the implementation and enforcement of the AI Act and to ensure that the new rules contribute to the development of the digital sector in Europe.