Access to data for training algorithms in machine learning: copyright law and right-stacking
Inger Berg Ørstavik · Edward Elgar Publishing eBooks · 2022
News services and information gathering services are increasingly relying on automated processes. As information sources are unlimited and often unreliable, it is important from an information security perspective that services using artificial intelligence to process information are able to draw and present reliable conclusions. This Chapter examines if and how exclusive rights in copyright law restrict data access for training algorithms for the purpose of machine learning. Because access to huge amounts of data is necessary to train algorithms to develop high quality artificial intelligence services, the rights of authors to individual works as well as rights in collections of works - database rights and publisher's rights - must be cleared. The Chapter discuss the role of copyright law in striking a balance between right holders and AI developers and how public interests can be included in the balance. The discussion plays into the larger policy debate: whether the role of intellectual property rights is developing from incentive mechanisms to control mechanisms.