ABSTRACT
There is a de facto double standard in copyright eligibility. Despite the insistence of the US Copyright Office (Office) that copyright law is applied consistently to all subject matter, in practice the user of AI-generated images needs to meet an unattainably high standard. This includes that the author had a fully formed conception and total control over the creation of the work, beyond time and space, which this author calls the ‘platonic’ ideals of copyright. The authors of traditional works, such as paintings and photos often do not have to meet this standard.
This article contrasts the Office’s rejections of copyright protection for AI-generated images with the Beijing Internet Court’s approval, highlighting a global debate on copyright eligibility to AI-generated products versus traditional expressive works. The article is the first one that critiques both stances for overlooking creativity’s inherent unpredictability, authorial spontaneity, and the notion that at some level fine-grained ideas become expressions, and advocates for a policy that counterbalances human and AI contributions to artistic, literary, and musical works.
Instead of replacing the ‘double standard’ with a unified standard, this article proposes a dual standard: one for human-created works; and a different one for AI-assisted products. Therefore, it is important that artists disclose the part of the work generated by AI. But equally significant is that providers of AI-services make a database of generated AI-products available to the Office, so that it will be able to compare the applications for copyright registration with those products generated by AI, to see whether the human intervention meets the threshold and originality. Until this is possible, there should be a moratorium on the protection of AI-generated products via copyright law or a sui generis right. In turn, and to balance the promotion of innovation and creativity, the Office should make available registered copyrighted works and the metadata of their authors that can be used as training data for AI-service providers, so that they have the metadata to compensate these authors. This author recommends preferential treatment to human authors to avert or at least slow down the dilution of human culture.
Friedmann, Danny, Creation and Generation Copyright Standards (February 24, 2024), New York University Journal of Intellectual Property and Entertainment Law, forthcoming 2024; Peking University School of Transnational Law Research Paper Forthcoming.
Leave a Reply