The era of non-committal AI experimentation in the legal sector is officially over. On January 13, 2026, the Supervisory Board (College van Toezicht) announced that this year’s office inspections by local deans will explicitly focus on the safe and diligent use of legal AI tools. A clear pattern has emerged from this first round of audits: while law firms are rapidly adopting the technology, they rarely maintain an auditable record of their verification processes.
This lack of demonstrable oversight creates an operational blind spot and poses significant disciplinary risks for legal professionals.
The Gap Between Policy and Practice
The NOvA guidelines on AI in the legal profession establish clear, inescapable boundaries for the Dutch Bar. Core values such as confidentiality (safeguarding attorney-client privilege), competence, and impartiality remain paramount. In practice, however, translating theoretical firm policies into robust, operational IT safeguards often falls short.
During an inspection, the local dean (deken) evaluates not just whether a written AI policy exists, but how it is enforced daily. Two critical risk areas stand out:
- Breach of Confidentiality: Feeding client data into public AI models without strict pseudonymization directly violates professional secrecy. Since this data can be used to train external models, it constitutes an irreversible data breach.
- The Verification Duty for 'Hallucinations': AI models continue to generate highly convincing yet entirely fictional case law. Lawyers who blindly incorporate these fabrications into their files risk severe disciplinary sanctions. As we recently analyzed in our article on AI in personal injury law, the Dutch Bar Association requires lawyers to thoroughly verify all generated output and retain personal accountability for every legal document.
The Importance of Auditable Processes
The most critical takeaway from the dean's findings is that a manual 'human-in-the-loop' check is absolutely essential—but verbally promising it is no longer sufficient. Law firms must make their AI verification processes auditable. If an inspector queries the oversight process for a specific case, the firm must be able to prove that the source material originated from a closed system, that ECLI numbers are directly linked to the official Open Data of the Judiciary (de Rechtspraak), and that the lawyer made the final independent legal assessment.
LEO: The Compliance-Conscious Choice
For law firms aiming to innovate without incurring unacceptable compliance risks, relying on standard, consumer-grade AI is simply too dangerous. PrudAI developed LEO, an 'enterprise' AI assistant built from the ground up to meet the stringent demands of the legal profession.
LEO guarantees confidentiality within a hermetically sealed cloud environment (where no data is ever used for model training) and forces fully transparent source citations for every piece of output. This enables lawyers to confidently prove which real documents and court rulings form the basis of a draft advice, instantly satisfying the auditable verification duty.
Would you like to know how LEO prepares your firm's administration and IT infrastructure for inspections by the local dean, ensuring safe operations within the NOvA guidelines? Please contact our specialists for a demonstration.