One in four questions in Dutch parliament “written using AI”

A study by Dutch broadcaster NOS found that approximately one in four parliamentary questions posed by MPs in the Dutch parliament were drafted using AI tools like ChatGPT or Claude. Using the AI detection tool Pangram, NOS examined 28,000 parliamentary questions and identified 26% that displayed AI-style characteristics. The prevalence of AI-generated questions surged to over 30% just before the summer recess, but declined to around 17% in July, nearly the same as the previous year.

Some MPs interspersed a sentence or two of AI-generated text into their questions, whereas others used the technology to craft entire strings of parliamentary inquiries. Dutch Prime Minister Mark Rutte faced criticism last week after it was revealed that about 30% of his social media posts contained AI-generated content. This included personal reflections on sensitive topics such as the mistreatment of Moluccan war veterans in the 1950s, for which he later apologized on behalf of the government.

Rutte defended his use of AI, stating that "like three-quarters of Dutch people I sometimes use AI to make things better." While some experts believe that the use of AI to formulate parliamentary questions is not inherently problematic, they caution against its use as a shortcut for in-depth research. AI developer Maarten Sukel argued that if MPs avoid using AI entirely, it would be a cause for concern.

However, he emphasized that using AI should not replace thorough research when preparing for parliamentary debates. MP Daniël van den Berg of the opposition party JA21 supported the use of AI, stating that it saves time and improves the quality of research. However, Marleen Stikker, founder of De Digitale Stad, an early online network in the Netherlands, expressed concern that using AI in parliament undermines the nation's democracy.

Stikker argued that outsourcing the democratic system to American AI companies' algorithms, which are not neutral, is absurd and unacceptable.