Artificial intelligence: can AI tools infringe copyright?
ChatGPT, Copilot, Dall-E: artificial intelligence-based programs can be useful, but from a legal perspective, they raise new questions. Can AI tools infringe copyright? Sabrina Konrad, Deputy Head of Copyright and Related Rights Legal Services at the Swiss Federal Institute of Intellectual Property (IPI), shares her opinion.
Sabrina Konrad: In Switzerland, copyright protection only applies to works created by humans. Now if someone uses ChatGPT merely as a tool, but performs the creative effort themselves, the output could be protected by copyright. You could compare it to a photographer using a camera. However, if the creative effort is performed by ChatGPT, then it has not come from a human and therefore cannot be protected by copyright as such. But there is something else to consider: if ChatGPT uses copyright-protected content from other works in its output, then these parts of the output are still protected.
Things can get complicated if ChatGPT uses content from a work. If this content is still recognisable in ChatGPT’s output, the author’s permission is generally required. Alternatively, if the source is indicated, a quotation could be used. Essentially, however, there is a real risk of copyright infringement. Anyone who uses copyright-infringing output from ChatGPT is also committing copyright infringement themselves. The infringement still applies.
There’s no consensus on that yet. Some argue that using content to train AI is not subject to copyright law because it does not ultimately lead to human perception of a work. There are, however, ongoing legal proceedings on the matter in the US and in England. We therefore have to wait and see what the courts decide and on what grounds.