The European Union has published the EU AI Act in its Official Journal, marking the beginning of a new era for artificial intelligence regulation in the bloc. The law will come into effect on August 1, with all provisions fully applicable in two years. However, some measures will be implemented sooner, with key milestones set for the coming months.
In six months, the EU will ban the use of certain AI applications, including social credit ranking systems, facial recognition databases, and real-time emotion recognition systems in schools and workplaces. The ban aims to protect individuals from potential harm and misuse of AI technology.
Nine months from now, the EU will introduce codes of practice for AI developers, with the EU AI Office working alongside consultancy firms to draft the guidelines. The codes will also apply to companies providing general-purpose models deemed to carry systemic risks, raising concerns about industry influence over the regulatory process.
After a year, makers of general-purpose AI models, such as ChatGPT, will be required to comply with new transparency standards and demonstrate the safety and explainability of their systems to users. The EU AI Act also includes rules for generative AI and manipulated media, ensuring that AI-generated content is clearly labeled and transparent.
Companies training AI models will need to respect copyright laws, unless their model is created solely for research and development purposes. The Act states that rightsholders may reserve their rights to prevent text and data mining, unless it is done for scientific research purposes. If the rights are reserved, providers of general-purpose AI models must obtain authorization from rightsholders to carry out text and data mining over such works.
