Artificial Intelligence and Copyright – Part 1: The protection of AI works

The impressive development of artificial intelligence in recent years raises new questions in every field. One of the most affected areas, also thanks to NFT and the emergence of digital art, is the market for art.

Leaving aside the question of the protection of these works on the basis of copyright and the ownership of these possible rights for a while, it seems reasonable to agree that artificial intelligence is now able to create genuine original works of art, perhaps not in the legal sense of the term, but at least in the sense of common language.

Works created by artificial intelligence also seem to have a market. Just think that in 2018, Christie’s auction house put up for sale a work created by an artificial intelligence, which then sold for around $430,000.[1]

Having ascertained the existence of genuine artificial intelligence creations, two questions need to be answered: are artificial intelligence creations protectable by copyright? Who is the legitimate owner of the copyright on these creations?

In this first part, we will try to answer the first question:

“Are artificial intelligence creations protectable by copyright?”

The prevailing line of thought is that artificial intelligence creations are not eligible for protection because they are not the result of a human creative process.

In this regard, we believe it is necessary to make an initial distinction between works created with the assistance of artificial intelligence (1) and works created autonomously by artificial intelligence (2).

  1. The protection of works created with the assistance of artificial intelligence

In the creation of these works, human intervention, the author’s artistic choices play a fundamental role.

In our opinion, the current legal framework is perfectly suited to accommodate and protect this type of creations.

Indeed, despite the intervention of artificial intelligence, after all, these creations are the result of human creative choices that reflect the author’s personality and, as such, deserve protection.

To cite a real-life example that a court in Beijing was confronted with, a person had attached a sports camera to a hot-air balloon and, upon releasing the balloon, the camera automatically took pictures of the surrounding area and then autonomously selected the best shots.

In discussing the issue of copyright protection of these photos, the court ruled that although the camera was out of human control during the process of shooting images automatically from the sky, there was decisive human intervention characterized by the selection and choice of factors such as camera type, positioning, camera angle, video recording mode, video display format, sensitivity.[2]

As long as technology does not evolve to such an extent that machines and systems, including AI, can be fully autonomous and operate independently, there seems to be no need to adapt the current copyright law system.

However, the issue changes with the likely progressive reduction of human intervention, a reduction that is already underway and will lead to autonomous or semi-autonomous creations by artificial intelligence.

  1. The protection of AI autonomously created works

As mentioned above, in most countries, the current legal framework cannot protect autonomous creations of artificial intelligence due to the absence of a human creative process.

However, artificial intelligence is evolving very rapidly, and we may soon be faced with the problem of protecting a large number of autonomous creations.

To cite an example, the language model called Generative Pre-trained Transformer (GPT), owned by OpenAI, a US non-profit company co-founded in 2015 by Sam Altman and Elon Musk, once it has received the huge masses of documents on which to base itself and a textual input, is able to generate images (but also text) without the need for further instructions.[3]

This model, entered with the input “an avocado-shaped chair”, produced the following creations:[4]

A picture containing green, indoor, room, yellow

Description automatically generated

We therefore believe it is necessary to reflect on whether or not a new legal framework should be built to protect these creations.

We cannot but agree with leading authors that today’s artificial intelligence can no longer be regarded as equivalent to the thinking and deduction processes of the human brain, but rather as the ability of machines to learn, through modelling, error and optimisation processes, by processing huge volumes of data.

However, in our opinion, the requirement of ‘human creation’ necessary for protection, is on the one hand a legacy of the historical period of the formulation of the body of copyright law, and on the other hand, will gradually become more and more anachronistic over the years as technology, artificial and human intelligence increasingly merge in the creative process.

Moreover, the mere fact that artificial intelligence lacks humanity does not, in our opinion, imply that its creative process cannot deserve protection. Triggering a debate about which creative or logical processes, which brains deserve protection based on the definition of humanity sounds us rather sterile and dangerous. Indeed, copyright protects every original work, whether it is the result of a human creative process be it rational or instinctive, conscious or unconscious.

Why would works that are unquestionably new, created through a process of selection, modelling, error, optimization not be worthy of protection simply because they do not originate from a human being?

Perhaps the answer is not to be sought in the law but rather elsewhere. After all, law is a tool to achieve a social goal, a tool which shows the way, once the goal has been established.

Artificial intelligence creations should be worthy of protection through copyright or through a sui generis right such as those specifically created for databases if this protection meets social and economic needs and purposes.

On the one hand, the absence of protection could provide a lot of material that can be freely used by anyone, on the other hand, it could discourage the development of these types of algorithms. The choice is up to us.


[2] Beijing Intellectual Property Court (2017) Jing 73 Min Zhong No. 797 Civil Judgment. April 2, 2020.