‘NYT vs OpenAI, Microsoft’ will set a precedent for copyright protection of creative work used for AI
This year has seen an unprecedented number of unexpected cover versions uploaded on YouTube: Freddy Mercury singing 'Sweet Child O' Mine' by Guns N' Roses, Johnny Cash singing Aqua's 'Barbie Girl', and even Frank Sinatra singing Coolio's 'Gangsta Paradise'. What's common among them is that all these singers are no longer alive.
These cover versions were never sung by them — they've all been generated by AI.
Copyright was created to ensure that creators have incentive to do new work. But it also leaves a little wriggle room for derivative work as 'fair usage': for criticism, comment, news reporting, teaching, scholarship or research, among others, where only a small sample may be duplicated.
For example, facts cannot be copyrighted, but a sequence of words, in terms of how those facts are expressed, are protected.
You can cite a part of the text, but only as much as may be seen as fair. However, who owns the copyright of an AI-generated work, if the music and lyrics are new, even if the audio is a mimicry of a famous singer?
AI is testing the boundaries of copyright by cloning music, voices, visuals and people, and generating new lyrics and scripts, based on ingesting previous work from writers, musicians and composers.
