What Does George R. R. Martin Have to Do with AI and Copyright?

The second post in our blog series on the possibilities of AI within a secure legal framework focuses on the topic of copyright. Where does AI get its data from, and what impact does this have? You can read about all this and much more here in the blog.

Data – The Raw Material of Functioning Artificial Intelligence

When training an AI, i.e., importing data, all available data is used and imported into the system. After all, data is the raw material for functioning artificial intelligence.
However, there is one important point to note:
The results that the system ultimately produces are nothing more than the results of other natural persons compiled by a natural person. These results are therefore not the outcome of complex AI work, but merely a compilation of existing knowledge. And this is where the problem lies: the final results produced by the AI are the work of a natural person who has painstakingly sat down and obtained these results through long hours of work. The results therefore belong to someone else: the author. The author is thus the intellectual owner of a work.

What, or rather Who, Does Copyright actually Protect?

Copyright protects the natural person who created a work. This can be a painting, a written document (e.g., a book), a photo, a film, or a piece of music, etc. The person (author) does not have to register their work separately (unlike with a patent or trademark); instead, the work is already protected once it is created. Put simply, copyright is a reward for the person who created the work.
The author can grant third parties so-called exploitation rights, i.e., third parties may publish, reproduce, or edit their works. In practice, this is often referred to as licensing. However, this must be done by the author themselves.

Is AI a Creator?

No, AI can never be an author, because AI is only as smart as the data it has been fed. Consequently, AI is limited in its knowledge pool, depending on how up-to-date the data sets are. AI cannot (yet) think independently and create its own results. It is nothing more than a database that the underlying system draws on.
It should be noted that AI itself cannot be protected by copyright. Only its source code as such can be protected by copyright. The results of AI can only be protected by copyright if a natural person edits them and thus creates new works.
However, copyright plays an important role in the development of artificial intelligence.

Data Mining to Feed the AI

The AI system is fed by means of “data mining.” This means that various data is collected, compiled, and processed. Various sources are used for this purpose (the Internet, studies, individuals, etc.). This can become a problem if the copyrights of the authors are disregarded and any necessary rights of use or licenses are not obtained.

And What Does George R. R. Martin, Author of “Game of Thrones,” Have to Do with all this?

The developers of ChatGPT (Open AI) used the Game of Thrones books without the author’s consent and trained ChatGPT’s AI with this data. Open AI fed ChatGPT with texts found online to train its language model. George R. R. Martin and other well-known US writers have now filed a class action lawsuit against Open AI for copyright infringement. The lawsuit alleges that “grand theft” has been committed.
The authors fear for their livelihoods, as such AI can create texts “cheaply and quickly” that would take a writer several months or years to produce, and for which they would naturally be paid.
(Source: https://www.tagesschau.de/ausland/amerika/klage-schriftsteller-chatgpt-100.html)

What Can We Learn from this Case?

In summary, it should be noted that copyrights must be taken into account when developing, and in particular when training, artificial intelligence. Data must not be imported arbitrarily; instead, potential copyright infringements must be examined and the rights of the copyright holders must be obtained in advance in order to avoid any violations.
Next week, we will continue with the topic of patent and trademark law and the extent to which these are relevant to artificial intelligence.