AI Art: What Can Machines Own?
In 2011, British photographer, David Slater, left his camera unattended on a tripod whilst in an Indonesian national park. Subsequently, a monkey named ‘Naruto’ spotted its reflection and took a picture of itself wide-eyed and smiling at the camera.
This led to a series of copyright disputes. In 2014 Wikipedia used the image on its site, claiming the photograph was in the public domain. Slater attempted to have it pulled for copyright infringement. In 2015, a lawsuit was also levelled by the People for the Ethical Treatment of Animals (PETA) arguing that the monkey autonomously produced the photograph, depriving Slater of any rights.
Whilst this case was settled outside of court in 2016, it raised questions as to what dictates human authorship in US copyright law. This is particularly relevant in the age of artificial intelligence (AI) art, where machines created by people learn from others’ art to produce unique images.
What does this mean for AI art and its copyright protection?
In recent years, AI art has grown in significance, both culturally and economically. It has taken leaps and bounds in its ability to replicate human themes and art styles through machine learning.
This growing market has become particularly significant for owners of art-generating machines due to potential profits. One of the most popular engines, OpenAI’s DALL•E, bills users $15 for 115 inputs into its system.
OpenAI, co-founded by Elon Musk, promotes its ability to use algorithms to assist in generating work that replicates the styles of well-known artists such as Salvador Dali. OpenAI is based in the US where currently there is no legislation that explicitly controls the collection or generation of AI data. For some this is troubling as it means that any work accessible on the internet, regardless of its legal protection, can be used to train AI. It can be argued that this creates opportunities for bad actors to imitate others’ work and profit from it, eliminating the need for originality.
However, a decision made by the US Copyright Board on the 14th of February 2022, seems to indicate that US litigators will be against granting copyright protection for AI art. The Board rejected a copyright application made by the AI artist, Steven Thaler.
Thaler used his own AI engine, the ‘Creativity Machine’, to produce “A Recent Entrance to Paradise”. He argued that he should enjoy the rights to the work as he was the owner of the AI and algorithms used to create it. This raises questions as to who should enjoy copyright protection and what ownership means in terms of computer generation.
The key issue was recognising ‘human authorship’ within the framework of US law. Chapter 300: 21-22 of the Compendium of the US Copyright Office Practices states that work cannot be copyrighted ‘without any creative input or intervention from a human author’. This implies that prompting AI, and collating sources from which it can learn, are not sufficient for human authorship.
Whilst not an act of government, or official ruling, this remains significant in light of the recent advances of OpenAI’s other software, ChatGPT, which is capable of creating lyrics, poetry and short stories.
Conclusion
As AI becomes increasingly significant, appearing to take over creative endeavours and diminish the need for commissionable artists, the law will have to adapt to determine what deserves protection.
Currently, the biggest players are based in the US, where IP law is unlikely to offer protection to the products of AI. This debate will continue to change and evolve as other countries begin to explore what role AI should play and how their laws should manage it.
By Samuel Cozier Axford