Legal edition
AI does not only affect case files. It can also influence how judges assess them.
UvA: AI can quietly shape what judges notice
An AI tool can influence a case before a judge has ruled on it.
Researchers at the UvA warn that AI can shape what judges pay attention to. A system can make some details stand out more and others less. This can strengthen existing biases. The researchers say it is not enough to say that AI only supports the judge. This matters to lawyers and other legal professionals. They want to know how AI affects the justice system. There are plans to use AI to review case files and draft judgments. A draft judgment is an early version of a ruling. AI could then have an influence before the judge writes their own decision. The researchers point to a trial in 2025 at the Rotterdam District Court. AI helped write a criminal judgment there.
In a case, find out what role AI may have played. If needed, ask how the judge checked the outcome. An AI system can decide which documents or details get extra attention. This means the person reviewing the case does not start with a neutral pile of information. Does the system show certain patterns more often? That could strengthen existing assumptions. So AI’s influence is not limited to its advice. The order and choice of information can also shape the decision.
The researchers say the word ‘supportive’ does not protect against this influence. The source does not set out a new rule for judges. Nor does it describe a fixed test for every case. Legal professionals should not assume that a final decision by a human removes all AI influence. The practical question is: which information did the system select, and how was that choice checked?
Background for subscribersThe reports show that AI can affect how information is selected, organised or assessed. So when it comes to the justice system, it is useful to ask two questions. What role did a system play? And how did a person check the result? The same applies to AI used in recruitment, healthcare or regular legal work. The data used, the testing approach and human checks all matter. The reports do not create new Dutch rules. So always check what applies in the specific case or organisation.
US court: this AI training did not qualify as fair use
A US court ruled on a case involving Ross Intelligence. The company trained an AI search engine using protected headnotes from the legal database Westlaw. The court said this was not allowed. It is the first appeal court ruling on AI training and fair use. The court stressed that the case was not about generative AI. That is AI that creates new text or other content. Lawyers expect parties to cite the reasoning in other ongoing cases too. These include disputes about music. The ruling raises a question for legal organisations. Where does the material used to build AI systems come from? The ruling does not automatically decide how other cases will turn out. But it shows how a court may assess copyright in training material. Check what an AI provider says about where its training material comes from. The court looked at protected headnotes from Westlaw. They had been used to train a competing AI search engine. The question was whether this use counted as fair use. This is a US exception to copyright law. The court found that the exception did not apply here. Training an AI system is therefore not automatically outside copyright law. The case was not about a generative model. That is a system that creates new text. A ruling by a US court also does not automatically set a rule for Dutch cases. Other parties may still use the reasoning in ongoing disputes. So carefully check what the ruling means for a contract or case under European or Dutch law.
Background for subscribersThis is a taster.
Subscribers read the whole Legal edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Legal edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026