The European Union has published the full and final text of the EU ai Law on its Official diaryas reported by ai-act-gets-published-in-blocs-official-journal-starting-clock-on-legal-deadlines/” data-ylk=”slk:TechCrunch;cpos:3;pos:1;elm:context_link;itc:0;sec:content-canvas” class=”link “>TechnologyCrunchThe new law will come into force 20 days after its publication, so it will be applicable from 1 August. All its provisions will be fully applicable within two years, but some of them will be implemented much sooner.
Within six months, the bloc will begin implementing bans on prohibited ai applications such as the use of social credit rating systems, the collection and compilation of facial recognition information for databases, as well as the use of real-time emotion recognition systems in schools and workplaces.
In nine months, the EU will begin enforcing codes of practice for ai developers. The EU ai Office set up by the European Commission will work with consulting firms to draft those codes. It also plans to work with companies that provide general-use models deemed to carry systemic risks. However, as TechCrunch notes, that raises concerns that major industry players could shape the rules that are supposed to keep them in check.
After a year, creators of general-purpose ai models like ChatGPT will have to comply with new transparency requirements and prove that their systems are safe and easy to explain to users. On top of all that, the EU ai Law includes rules that apply to generative ai and manipulated media, such as making sure deepfakes and other ai-generated images, videos, and audio are clearly labeled.
Companies that train their ai models will also have to abide by copyright laws, unless their model is created solely for research and development. “Rights holders may choose to reserve their rights to their works or other subject matter to prevent text and data mining, unless this is done for scientific research purposes,” the text of the ai Act says. “Where opt-out rights have been appropriately expressly reserved, providers of general-purpose ai models must obtain authorization from rights holders if they wish to perform text and data mining on such works.”
