AI Models Face Legal Quandary Over Copyrighted Training Data
-
Generative AI models like ChatGPT are accused of copyright infringement for training on others' creative work without permission. Media companies are suing and striking deals.
-
Models argue their use is "transformative" and constitutes "fair use", but laws vary globally. Rights-holders want compensation for providing AI's raw materials.
-
Some creators like newspapers and Getty Images are developing their own AI tools trained only on their content, retaining control and revenue.
-
Deals so far provide "anaemic" royalties to publishers. Building own AIs risks being outcompeted by giant general models.
-
Laws and rights around AI creations are unclear. Works made solely by AI may not be copyrightable at all, foreshadowing a world of unownable content.