Dutch Parliament AI Use in Speeches

That raises questions about transparency in government at a time when public trust is already under strain. If legislators are relying on chatbots to compress arguments, polish phrasing or generate drafts, the line between original political judgment and machine-assisted output becomes harder to see. In a legislature where speaking time is limited and messages are increasingly optimized for brevity and clarity, AI is becoming a productivity tool — but also a potential credibility risk.

The investigation, which ran the speeches of all 150 lawmakers through Pangram, suggests the practice is not confined to a narrow group. While the top of the list included lawmakers from across the spectrum, the analysis found overrepresentation among far-right factions such as the PVV and Groep Markuszower. By contrast, the SP, Party for the Animals and Volt were not found to use chatbots in the sampled speeches.

The use appears most common in the less visible parts of parliament. The report said AI was seldom detected in major plenary debates involving party leaders, but in committee sessions it sometimes accounted for more than half of opening remarks. That distinction matters: committee work is where much of the legislative detail is hashed out, and where staffing constraints can be tighter, especially for smaller parties and newer lawmakers.

Several members acknowledged the practice, but framed it as editing rather than ghostwriting. PVV lawmaker Van Meetelen said he would never deliver a speech he did not stand behind. BBB leader Caroline van der Plas defended limited use of AI as a practical tool for a small faction with heavy workloads. CDA member Derk Boswijk said he was surprised to rank so high and warned against the impression that his speeches “roll entirely out of Chat.”

The political sensitivity is obvious. Party leaders such as CDA chief Henri Bontenbal are already pushing for internal rules, underscoring that the issue is shifting from novelty to governance. Earlier reporting had already shown that Prime Minister Jetten regularly used AI for social media posts, and NOS found last month that a quarter of written parliamentary questions were prepared with AI assistance.

For investors, the broader significance lies less in the Dutch parliament itself than in the fast-growing demand for generative AI tools across regulated, reputation-sensitive industries. Microsoft and Nvidia have both disclosed in filings that AI use creates regulatory, legal and reputational risks, including scrutiny over misinformation, privacy and political misuse. As governments draw sharper lines around acceptable use, vendors of AI software and infrastructure may face tighter compliance demands even as adoption widens.

The likely next phase is not a retreat from AI, but a push for disclosure rules, internal standards and clearer boundaries around what counts as acceptable machine assistance. For lawmakers, the key risk is erosion of authenticity. For AI providers, it is that political use cases become another front in the regulatory debate over whether these tools are augmenting public discourse or quietly rewriting it.