The imperfect labeling of AI-generated content

The decisive question is not simply whether artificial intelligence generated a text. It is whether the person to whom the text is attributed has the intimate relationship with it that underpins authorship, writes Baptiste Le Bihan, a philosopher and assistant professor in the Department of Philosophy at the University of Geneva.

From Aug. 2, 2026, the European Union’s artificial intelligence regulation is due to require content produced by generative AI systems to be identifiable, including through digital watermarks. Such labeling would indicate that AI generated the text.

But that signal would not, by itself, establish whether the generated content contains an AI contribution of its own or instead reflects an original human contribution. It therefore cannot alone determine whether the human involved supplied the creative contribution that justifies authorial status.

The distinction between generation and authorship is not entirely new. The article continues by placing the question within the longer-standing issue of what makes someone the author of a work.