The protectability of machine learning models

Reflections on the interface of law and technology

The potential of artificial intelligence(AI) seems limitless. One sub-discipline of AI is machine learning(ML). At the very heart of ML are the so-called machine learning models(ML models), which store knowledge acquired through training. The economic value of ML models is correspondingly high, and with it the interest in their legal protection.

The starting point of the dissertation project is a definition of the often technically circumscribed and imprecisely used term ML model and its differentiation from other terms. Based on this, it is examined how ML models are protected de lege lata against the background of copyright, patent, unfair competition and contract law. This is important because ML-models are a complex subject of protection and cannot easily be classified into the traditional categories of copyrightable works or patentable inventions.

The aim of the study is to develop a comprehensive standard work on the protectability of ML models in Switzerland, which creates more legal certainty, identifies regulatory gaps and thus also proposes sustainable solutions de lege ferenda.

Prof. Dr. Alfred Früh is supervising Noëmie Schär's dissertation project. The study is financed by the Swiss National Science Foundation through a Doc.CH grant.