Contents
Many legal questions about artificial intelligence are still unresolved. An overview of the largest construction sites.
Artificial intelligence that draws pictures or writes texts on command has been making headlines for weeks. However, there is seldom any mention of legal problems arising from the use of such systems – only one thing is certain: such questions will pose new challenges for the courts in the years to come.
A lot about the use of these technologies is still unclear today. After all, it can be said where the biggest construction sites are:
Construction site 1: Copyright
In the training material of AI systems, there are always copyrighted images or texts. The machine can later generate content based on the training material itself. Was she just being inspired or was she infringing on copyright?
At the request of SRF, the lawyer Martin Steiger, who specializes in law in the digital space, can only state: “Whether there is a copyright infringement in an individual case, the copyright must clarify. Such legal disputes are often complex and lengthy.”
Construction site 2: ancillary copyright
The training material of AI search engines also includes the content of publishers and news agencies. This raises the question of whether publishers can assert an ancillary copyright against the operators of AI search engines – there is no answer to this yet.
Emanuel Meyer from the Swiss Federal Institute for Intellectual Property says: “The Federal Council will shortly present a draft law for a publisher’s ancillary copyright. This one is not tailored to AI search engines. A corresponding expansion is likely to become the subject of further political discussion”.
Construction site 3: data protection
Texts of a personal nature can also be included in the training material of AI systems such as ChatGPT. It is not possible to check what data has been collected from you and there is no way to request that this data be deleted. It is therefore controversial whether ChatGPT even meets the requirements of the European General Data Protection Regulation.
Suzanna Marazza from “CCdigitallaw” and the Università della Svizzera italiana explains: “If the data provided by the users are processed in any way or – in a non-anonymised form – collected as training material, the law on data protection applies.” However, in practice it is difficult to enforce the rights of those affected. For example, when it comes to deleting data that has already been collected.
Construction site 4: Moderation of the content
The providers of AI systems try to prevent their artificial intelligences from outputting problematic content – nevertheless, this has happened again and again in the past.
The lawyer and media and copyright expert Simon Schlauri: “The results of artificial intelligence are not fully controllable. If something happens in an individual case, although the provider has taken appropriate protective measures, you can hardly blame him.»
The discussion is likely to be going in the same direction as in the field of autonomous driving, Schlauri continues: It cannot be demanded that AI systems function 100% error-free.
Date vous, February 24, 2023