With the spread of ChatGPT, artificial intelligence is now on everyone’s lips: from texts to images, the outputs of this new technology amaze us and open the door to many possible applications.
In addition to the debate on the quality of the content and the ethical implications arising from the use of these tools, there is another important aspect to consider. Are artificial intelligence creations covered by copyright? Or rather do they infringe the copyright of other works produced by human beings? Let’s try to see clearly!
How AI generates text and images
Before delving into the issue of copyright related to artificial intelligence, let’s take a small step back to understand how AI (Artificial Intelligence) works.
The different platforms, both those for generating texts (such as ChapGPT ) and those used for creating images (such as DALL-E ), are based on machine learning algorithms, systems capable of learning automatically starting from already existing data.
Applied to AI, this means that software is trained with content previously produced by humans to create images and texts. To simplify, we can say that they rework the information available to them to develop new content.
On a practical level, using AI platforms is really simple: a short textual input is enough to obtain the desired result. For example, if you enter “Write an article on copyright and artificial intelligence” in ChatGPT you will have a text that talks about this topic (don’t worry, the article you are reading is the result of totally human intelligence!). Likewise, if on DALL-E you type “child running on the beach on a sunny day” you will get what you hope for (more or less).
Are images and texts created by AI protected by copyright?
Now that you know how this software works, let’s go back to the first question with which we opened the article. Are artificial intelligence creations protected by copyright? The answer is no. Or rather not at the moment and not at all!
Given that this sector is still in its infancy but evolving very rapidly, regulatory issues are also in flux and dedicated laws will certainly be needed to regulate the matter.
AI outputs are not currently subject to copyright. This is because in the United States only works developed thanks to the creative contribution of a human being are protected. If the work was created using artificial intelligence software, this connection is lost and the human contribution is considered practically irrelevant compared to that of the machine.
However, the issue is more complex than it seems because AI can only be used by human artists as a support or stimulus for the creation of works. What happens in these cases? Everything seems to depend on the level of human contribution to the development of the text, image, or audio, which is not easy to establish.
Both the United States and the European Union seem to agree that works created by humans with the help of AI can fall within the framework of intellectual property protection.
Do images and texts created by artificial intelligence violate copyright?
We now come to the second problem, which arises especially when such creations are used for commercial or marketing purposes. Given that AI software is “inspired” (in part) by human works protected by copyright and starts from these to generate their output, does the development of automatic content violate copyright?
Here too the answer is not simple. On the one hand, the algorithms originally reprocess the data and do not produce an exact copy of the initial works: it would therefore be a new creation. In the United States, AI companies rely on the fair use doctrine, which allows copyrighted material to be used for fair purposes. On the other hand, humans have always taken inspiration from the creations of artists who came before them: why shouldn’t AI be given the same opportunity to be trained?
However, the question is controversial. The artists whose works are fed to machine learning models claim that they have never given consent for this particular use of their works. Furthermore, what happens if the AI copies the personal style that characterises a designer or writer?
Recently, three American illustrators sued several large image AI companies for using their creations to train software without the artists’ permission.
Even regarding the automatic writing of texts, the debate is open. If platforms fish for online content to develop new texts, how is it possible to verify the correctness and truthfulness of the information? What are the sources from which the software draws?
“Safe” artificial images the case of Shutterstock
An interesting case to consider from this perspective is that of Shutterstock. The famous platform, which provides photographic content and illustrations, has started a collaboration with OpenAI and LG AI Research and has included a new section of the site called Generate, which allows you to develop the desired image with artificial intelligence.
The interesting thing is that Shutterstock has declared that with this system “safe” images are created, i.e. without any copyright infringement. In this regard, CEO Paul Hennessy declared :
“Our tools are built on an ethical approach and a library of resources that represents the diverse world we live in, and we ensure that artists whose works have contributed to the development of these models are recognised and rewarded”.
The images that are created with the Generate function can be downloaded in high resolution with the license to use for commercial purposes.
Artificial intelligence a look at the near future
Having reached the end of this journey, it’s time to take stock. The future of artificial intelligence certainly remains to be written (or designed)… perhaps this time without the help of automatic tools!
The current situation makes it necessary to have a system of regulatory protections that regulates the applications and limits of AI, both for those who create content using AI as an artistic support and for those who ask that their copyright be respected.
At site By site, we carefully follow the evolution of discussion and technologies, and we experiment with new content creation systems considering all the aspects that are at stake.
