Bringing AI to the Hoge Raad in a secure and private environment with UbiOps

Generative AI for document processing and legal research in a private environment

Challenge

The Supreme Court of the Netherlands is bound by national security rules, requiring an on-premise AI solution that scales efficiently and ensures confidentiality, as public cloud services are not allowed.

Solution

UbiOps provides a secure and flexible infrastructure, enabling AI development while ensuring compliance with national security standards.

Result

The Hoge Raad is currently running multiple projects with UbiOps, which offers scalable AI solutions and secure GPU deployment without additional infrastructure.

Bringing AI to the Hoge Raad in a secure and private environment with UbiOps

Generative Artificial Intelligence (GenAI) is becoming an essential part of workplace efficiency. From summarizing and structuring texts to retrieving information, online AI models
are being integrated into daily processes everywhere. But what if your organization works with classified data governed by strict regulations? Closed-source AI solutions running in
public clouds are a definite no-go in such cases.

That’s the challenge the Hoge Raad faces. They want custom AI solutions to help their team work smart, but strict regulations require those solutions to stay on-premise with open-source models. Building the necessary infrastructure, including access to GPUs, would be both costly and time-consuming. Luckily, UbiOps provides a solution to this problem.

The Hoge Raad

The Hoge Raad der Nederlanden (Supreme Court of the Netherlands) plays a crucial role in maintaining consistency and uniformity in civil, criminal, and tax law by adjudicating cassation cases. Its primary function is to rectify legal errors, ensure the proper interpretation and application of the law, and establish binding precedents that shape future judicial decisions.

For many years, the research supporting their work relied heavily on extensive libraries filled with books, housed in quiet rooms lined with countless shelves. Today, however, the Hoge Raad operates a state-of-the-art digital knowledge center, containing thousands of ebooks and case files, all accessible via custom-built search engines. To safeguard the sensitive information they manage, all data remains strictly on-premises, ensuring the highest level of confidentiality.

The Importance of AI

AI has rapidly evolved in recent years, offering tremendous efficiency gains for many organizations. Adopting AI presents a unique challenge for Hoge Raad, which deals with classified information daily. Public AI models are not an option due to security concerns, and even though cloud solutions are often considered safe, regulations require that AI be implemented on-premise to ensure complete control over data.

“At the Hoge Raad, safeguarding highly sensitive information is a top priority,” said Michel Mooren, IT consultant at the Hoge Raad. “However, we recognize that AI has the potential to enhance our efficiency and accuracy, particularly in areas such as document processing and legal research.”

Pseudonymization with PseudoKeeper

Let’s take a look at the first project: PseudoKeeper.

Pseudonymization is essential for the Hoge Raad. It involves replacing sensitive information in legal documents with unique identifiers, ensuring compliance with data protection regulations. Traditionally, this process is carried out manually, a time-consuming and error-prone task, even when performed by highly trained professionals. The potential for AI to take over much of this work is immense.

With PseudoKeeper, an AI-powered pseudonymization tool, the Hoge Raad can automate most of the process. The tool handles most of the work, while trained staff only need to perform a final check. This is where UbiOps played a critical role: by using UbiOps’ Private cloud-based version to train the models with publicly available data, the Hoge Raad could take advantage of existing infrastructure, including access to GPUs, saving time and resources.

However, because of security rules, classified data could not be used in the cloud to train the model. Instead, the Hoge Raad used milestone legal cases predating 2000, publicly available on rechtspraak.nl, as training data. This allowed them to train PseudoKeeper without risking confidentiality.

“PseudoKeeper is going to save us so much time,” Mooren added. “Manual pseudonymization is tedious and prone to errors. With this tool, we’re automating much of the work, leaving our team to focus on the final review.”

Hoge Raad is now in the final stages of developing PseudoKeeper, and they
plan to integrate it fully on-premise for classified documents. UbiOps will
soon also be installed on their servers, enabling the Hoge Raad to maintain the same efficiency level while complying with all regulatory requirements.

The Legal Assistant with RAG Technology

The second project involves developing a legal assistant using RAG technology (Retrieval-Augmented Generation). This tool is designed to assist legal researchers by retrieving and summarizing legal documents. The assistant pulls from a custom-built knowledge base, generating clear, concise answers to complex legal queries.

This RAG-based legal assistant has the potential to significantly improve the efficiency of legal research at the Hoge Raad. By quickly retrieving relevant documents and summarizing them, the assistant could streamlinethe entire research process, reducing the time spent sifting through case law and legislation. While it’s still in the development stage, early iterations of the tool have already shown promising results.

Results and future outlook

One of the most significant advantages of UbiOps is its ability to run on-premise and in a private cloud while remaining easy to use. The Hoge Raad can develop robust AI solutions without investing heavily in new IT infrastructure or acquiring expensive GPUs. The platform provides the flexibility they need to operate within their secure environment while still
allowing AI to scale as needed. “As a small IT team, we rely on UbiOps to provide the infrastructure we don’t have in-house,” said Mooren. “UbiOps gives us the tools we need to build and test AI models without having to invest in specialized hardware or expertise.”

Looking ahead, the Hoge Raad plans to fully integrate the PseudoKeeper tool on-premise and expand the legal assistant to cover additional areas of law. By 2025, they expect to roll out even more AI tools, including an automatic case file summarizer, which will further enhance the efficiency of legal research.

“UbiOps is helping us stay ahead in a rapidly evolving field,” said Mooren. “With their support, we’re confident that we’ll be able to implement AI solutions that make our work faster and more efficient, all while keeping data secure.”

The UbiOps Advantage

The Hoge Raad’s collaboration with UbiOps shows that even the most highly regulated institutions can adopt AI while fully complying with data security and national regulations.
UbiOps offers the perfect balance between security, scalability, and AI innovation for organizations dealing with similarly strict rules.

Are you part of an organization that handles sensitive data and must comply with strict national security regulations? Contact us to see how UbiOps can fit into your development
plans.

Latest news

Turn your AI & ML models into powerful services with UbiOps