How AI is revolutionising work at the Federal Supreme Court – a podcast episode

Public authorities and the judiciary enjoy our special trust. The requirements for the use of artificial intelligence (AI) are therefore particularly high. Our AI expert Marcel Gygli has developed his own language models for the Federal Supreme Court. He heads the Digital Susatainability Lab at the Institute for Public Sector Transformation at BFH Business School.

The Federal Supreme Court issues a large number of judgements and leading decisions every year. These have to be published anonymously – all personal names, companies and places have to be made unrecognisable. Until now, this task has meant an immense amount of work. AI can make this much easier. If ChatGPT were data protection compliant, the prompt could be: anonymise this verdict. Of course, this is a no-go for authorities and the judiciary. But our AI professor Marcel Gygli, together with Joël Niklaus and his team, has trained a language model exclusively for the Federal Supreme Court that solves this task. To do this, the researchers trained the AI on historical data from the court. This worked so well that they have now started a second project. He talks about this and the ethical requirements in the new episode.

More on the topic:

• About the guest Marcel Gygli
• Publications by Marcel Gygli
• Dissertation by BFH researcher Joel Niklaus on the language model for the Federal Supreme Court
• Legal NLP – a BFH business competence

This podcast is produced with the kind support of: Audioflair Bern and Podcastschmiede Winterthur.