Dutch police test AI to crack down on environmental crime

The Dutch National Police and KickstartAI have begun user testing a new AI tool designed to help detectives tackle complex environmental crime cases. The prototype—working title “Al Capone”—promises faster discovery of relevant legal angles buried in sprawling, intricate case files. According to KickstartAI, the technology zeroes in on the earliest phase of an investigation, where teams lose the most time sifting through legislation, case law, and legal commentary.

The name Al Capone is deliberate. The infamous American mobster wasn’t brought down for his headline crimes, but for tax evasion. That idea anchors the project: don’t just chase the obvious offense—surface alternative legal routes that might still put suspects within reach of prosecution.

Environmental crime ranks among the most complex to investigate. Cases often contain vast volumes of unstructured data and cut across multiple areas of law. Beyond environmental regulations, investigators may face elements of fraud, money laundering, transport law, and administrative law.

KickstartAI warns that key legal leads are easily missed. Today, detectives manually comb through statutes, court rulings, and legal commentaries to map potential violations—a slow, expertise-heavy process spanning many legal domains.

The prototype ingests case documents and first produces a structured summary of the available information. It then charts a search strategy across four legal pathways:

Next, the AI queries preprocessed legal sources, including Dutch legislation, Dutch case law, European regulations, explanatory memoranda, and international jurisprudence.

Under the hood, it uses a Retrieval-Augmented Generation (RAG) architecture. Legal information is stored in a searchable vector database. A large language model then blends multiple search methods and ranks relevant laws, rulings, and legal angles based on the content of the case file.

The system does not make decisions for investigators. Human experts retain full control. The AI serves as a thinking partner—surfacing ideas and legal connections that might otherwise stay hidden.

Amir Niknam, Senior Innovation Advisor at the National Police, says the tech could be especially powerful in the opening phase of a case.

“Environmental crime often overlaps with other offenses and spans multiple legal domains. In the early stages, you need quick visibility into possible legal footholds. This test should show whether AI can help detectives outsmart criminals.”

The project builds on an earlier AI challenge, where police sought better ways to navigate complex environmental law and deploy the full set of legal tools more effectively against polluters.

The initiative reflects a broader trend: police forces worldwide are experimenting with generative AI and legal search systems, with a growing emphasis on augmenting human investigators rather than automating decisions.

Researchers note that AI can process large document sets, spot patterns, and reveal legal links faster. Still, human oversight remains critical to avoid errors, bias, and flawed conclusions. Studies on AI in law enforcement consistently stress: use AI as support, not as a substitute for human expertise.

The current user tests aim to gauge the prototype’s real-world value—what features actually help detectives, and what needs refinement.

After testing, the team will evaluate and iterate. By early August, the project is expected to deliver a recommendation for a minimum viable product (MVP), complete with documentation and a handover for potential further development.

For KickstartAI, this is a showcase of “applied AI”: not tech for tech’s sake, but targeted tools that help professionals tackle concrete societal challenges.