In response to a significant increase in submissions for copyright of creative content made at least partially with generative artificial intelligence (AI), The Copyright Office issued new guidelines in March 2023. Traditionally, human authorship has been requisite for applying for copyright. However, it is clear that artificially generated creative content will have increasing market importance, and as such the purpose of the recent guidelines is to soften the so-called “human authorship” requirement for copyright.
This poses something of a conceptual challenge, however. On one hand the copyright office wants to say that tools have always been used in the creation of copyrighted material, and thus the instrumental use of AI ought to be permitted as the paintbrush is to the artist, the typewriter is to the author, the instrument to the musician, etc. On the other hand, however, copyrighting material generated intrinsically by AI would violate the principals of copyrighting as such, namely to safeguard the intellectual property of citizens.
In an attempt to accommodate such concerns, the language of the guidelines allows for creative content to be copyrighted when it is deemed to have “sufficient human authorship to support a claim.” AI generated content that has been arranged or modified by a human in a “sufficiently creative way” would therefore be permissibly copyrighted. Applicants must disclose whatever AI generated content is involved in the material for copyright in the “Author Created” field of the application, and include a precise description of their contribution to the overall creative product. Furthermore, applicants who have already filed for copyright of AI generated materials must update their applications to include similar specifications about their contributions or risk losing their previously granted copyrights.
In spite of these guidelines, however, questions remain about how human authored content can be safeguarded in the age of generative AI. The threshold for what counts as “sufficiently creative” is left open by design, but it isn’t clear what about such a standard would prevent even extremely minimal human alterations to AI generated content to count as worthy of copyright. The issue of possible creative arrangement of AI generated content seems even less clear. If one were to produce, for example, a collection of paintings with seed input from Picasso, one could potentially claim that the mere arrangement of these images in a book merits ownership of the content? Examples like this clearly show the risk to artists, scholars, and other producers of creative content who may have their work effectively claimed without their consent by being filtered as seed content into generative AI programs.
Source: Artvy AI
On the other side, as AI continues to develop in sophistication, it may put the human authorship requirement completely into question. For one thing, the precedent for this rule appears to just as easily allow for a simple agent/patient distinction that wouldn’t be quite so anthropocentric, and according to which, any agent capable of producing content (human or otherwise) ought to be able to copyright material. Such developments would obviously have profound broader implications about the inclusion of non-human intelligence as possible bearers of rights and potential property owners. The accelerating trends in AI development also seem to indicate that such a future may not be far off and that more substantive modification of Copyright law will be increasingly inevitable.