Opinion: ‘AI must not take the place of the judge’

AI is making inroads into all sectors, including the judiciary. Is this a positive development? According to Benedikt Schmitz, it could save time. ‘But AI must not take the place of the judge.’

^{Text: Esther van der Meer}

Schmitz (31) is an Assistant Professor in Private International Law and Comparative Contract Law. The topic of the digitization of the judiciary is one he just happened to stumble upon. This spring, he organized a European conference dedicated to the digitization of the judiciary. One thing led to another, and together with two German colleagues, he recently founded the European Group on the Digitalisation of Justice, an intradisciplinary research consortium focusing on the digital transformation of legal systems across Europe. Its website has just gone live.

‘I take a fairly down-to-earth view of AI. I’m in the camp that sees it as a technological innovation, not a revolution that will eventually render humans redundant.’

But even as a technological innovation, there is still plenty to be cautious about. On 2 August, the next phase of the EU AI Act will come into force. It regulates AI systems based on the risk they pose to people and society. 'The judiciary falls into the highest risk category. The iura novit curia principle is essential: judges are expected to know the law and must ensure that the correct legal rules are applied. Blindly relying on AI is therefore dangerous, and automatically convicting someone without human intervention is completely unacceptable.'

Officially, therefore, AI is not yet used — or only to a very limited extent — in the judiciary. In practice, however, the situation is different. ‘You can already see lawyers using AI to brainstorm and even to draft petitions or summonses. And I wouldn’t rule out the possibility that judges occasionally use a chatbot too.’

Two years ago, a case before the magistrates’ court concerning a roof extension that was alleged to block sunlight from reaching the neighbours’ solar panels caused quite a stir. In the judgement, the judge openly stated that he had used ChatGPT as a source to determine the lifespan of solar panels, without consulting any experts.

But there are other ways to approach this. In Brandenburg, they are using specially developed AI to help judges answer very specific questions. ‘You can use it, for example, to quickly look up previous court rulings.’

This is where Schmitz sees potential. In the Netherlands, the number of court cases left pending is steadily rising, and there is a persistent shortage of judicial capacity. ‘If you can search through evidence much faster than at present, that could save time.’

There are also interesting developments possible in the field of cross-border law. ‘If, for example, you are a Dutch judge with a question about the Bulgarian legal system, it is currently expensive and complicated to find the answer. AI could be very helpful in that regard. But only under certain conditions: the AI model would need to be developed on behalf of the EU and fed with reliable data. Even then, however, there would still need to be a human being to check everything.’

And AI must never take the place of the judge, Schmitz believes. ‘One of the most important legal principles is that a judge must be able to explain how they reached their decision and what their reasoning is. That is not something AI can do. Perhaps AI will one day achieve the superintelligence that some are currently predicting. But as I said: I take a down-to-earth view on this and do not think it will happen.’

More information

• Benedikt Schmitz
• Overview of all opinion pieces