EU sets new AI regulations to enhance transparency and safety

In the European Union, new regulations concerning artificial intelligence are set to be implemented starting August 2, 2025. The EU Commission has recently issued guidelines aimed at ensuring transparency for providers of general-purpose AI models. These guidelines will require companies such as OpenAI and Google to disclose critical information regarding how their AI models function, the data utilized for training these models, and the measures they take to manage potential risks. The goal of this initiative is to prepare AI providers for the upcoming regulatory landscape. The upcoming regulations will specifically target AI systems capable of generating text, images, and videos, and these systems must be trained with significant computational resources. As part of this regulatory framework, extra obligations for risk assessment and mitigation will apply to particularly powerful models with systemic risks, affecting fundamental rights or security. The EU Commission's approach aims to clarify legal expectations, enabling providers to innovate while adhering to safety standards. The phased introduction of regulatory oversight will begin from 2026, granting the newly established EU AI Office the authority to review new AI models. Following this, starting in 2027, existing models will be subjected to similar reviews. These measures are designed to ensure ongoing compliance and to help mitigate potential risks associated with AI technologies. The Commission has communicated that they will work collaboratively with providers during the transition period before the regulations are fully enforced. Violations of the guidelines, once in place, could face hefty fines, emphasizing the seriousness with which the EU is addressing the intersection of technology and safety. This regulatory framework reflects the EU's commitment to fostering an environment of legal clarity that simultaneously promotes innovation in the field of AI whilst protecting the public and ensuring accountability from technology providers.