The non-committal era of artificial intelligence in the legal sector definitively ended in the late summer of 2026. Nine months after the introduction of the NOvA AI directives for the legal profession, these professional frameworks are no longer just tested on paper, but directly in the disciplinary court. Recent developments show unequivocally: copying an AI-generated summary of facts without meticulous source verification is now a direct disciplinary offense.
From policy to case law
When the Dutch Bar Association (NOvA) published its formal recommendations for AI use in late 2025, the core message was crystal clear: the lawyer is and remains ultimately responsible. In theory, the profession understood this, but translating it into fast-paced daily practice proved more challenging. In July 2026, we saw the first concrete signals from the disciplinary courts: submitting incorrect, hallucinated information generated by a generic AI model led to a formal reprimand for a Dutch lawyer.
This incident confirms an operational risk we already observed during the initial NOvA office inspections of 2026. Many law firms have embraced AI on the work floor but lack a foolproof verification process. When drafting a summary of facts or paraphrasing complex case law, standard LLMs fill gaps with plausible-sounding but incorrect details. Without auditable control mechanisms, this poses a massive risk for both the client and the firm.
The strict requirement of source verification
A lawyer has an independent duty to rigorously verify facts and legal sources. The NOvA recommendations dictate that legal professionals must exclusively use tools that are transparent and provide adequate source references. Yet, public or generic language models are still widely used, despite being unable to prove exactly where their conclusions are based.
For a future-proof law practice, this falls short. The Council for the Judiciary, which has been working with its own strictly controlled AI program since 2025, is also highly critical of the traceability of submitted documents. If you, as a lawyer, cannot immediately click on the original legal text or the specific paragraph from a previous ruling to substantiate an AI argument, the software is inherently unfit for your workflow.
Why LEO sets the operational standard
This is precisely why PrudAI made fundamentally different, safer choices when developing our AI Services for the legal profession. Our AI assistant, LEO, does not act as a blind text generator but is anchored in the principles of Retrieval-Augmented Generation (RAG), directly linked to current national case law and legislation.
Every legal claim, defense, and summary of facts that LEO generates comes with direct, clickable source references. This transforms the disciplinary duty of verification from an endless, manual search into an integrated and instantly auditable part of your workflow.
Would you like to know how LEO helps your law firm operate safely, efficiently, and fully compliant with the NOvA guidelines? Contact us today for a custom demonstration.