PAINTING A PICTURE OF COPYRIGHT SUBSISTENCE: ANALYSING ARTIFICIAL INTELLIGENCE-GENERATED PAINTINGS

Abstract

The advent of artificial intelligence (“AI”) has recharacterised the role of humans in a painting’s creative process. This recharacterisation introduces uncertainty to the Singapore Copyright Act. The extent to which the Copyright Act provides copyright protection for original computer-generated literary, dramatic, musical and artistic (“LDMA”) works is currently unclear. In particular, it is uncertain what is required to be “original” for computer-generated LDMA works under section 27. Regrettably, despite widely reported potential capabilities of AI in the 21st century, Singapore’s latest and extensive copyright review failed to consider this matter. In light of this and, additionally, given the sizeable investments pouring into AI systems that produce computer-generated paintings of substantial commercial value, this paper proposes a judicial test and statutory provision. As this paper treats “computer-generated” as the umbrella term for both “AI-generated” and “AI-aided,” and only the latter satisfies the Copyright Act’s originality requirement, the proposed three-stage sequential judicial test is to enable courts to distinguish AI-aided paintings from AI-generated paintings, by analysing the human input throughout the creative process. The proposed provision only awards copyright protection to “original computer-generated LDMA works,” and defines “computer-generated LDMA works” as “AI-aided LDMA works and not AI-generated LDMA works [emphasis added].”