AI-generated objections overwhelm Dutch municipalities
Local governments in the Netherlands are facing a significant rise in automated and AI-generated objections, overwhelming administrative systems. The objections primarily relate to the property tax system (WOZ), which is used by municipalities to determine property values for real estate taxes.
Businesses offering no-fee services had automated the objection process, allowing clients to submit objections by simply filling out an online form. This quickly led to a rise in costs for municipalities, as each objection can cost around 600 euros. The increasing number of objections overwhelmed staff and left boards and courts struggling to keep up with other matters.
Daniël Verloop, an AI specialist for the municipality of Montferland, said the volume of objections surged dramatically in late 2022 and early 2023. “We were flooded. That’s the best way to describe the increase in objections that came in,” he said. “A savvy entrepreneur developed an AI model to make mass objections, causing the number of complaints to rise by 400 percent in a short time.”
In response, a new law was introduced earlier this year to curb the profit-making model of these objection firms. According to the Dutch Valuation Chamber, the number of objections against property tax decisions decreased from 685,000 in 2023 to 325,000 in 2024, suggesting that the new regulation is having some impact.
However, a new challenge has emerged with the rise of AI-generated objections. While the volume of automated complaints has decreased, objections crafted by self-learning systems are now on the rise. De Telegraaf's survey showed that, out of 342 municipalities surveyed, 120 responded. Of those, 55 reported an increase in objections, and nine municipalities noted that some objections were generated using AI.
One municipality reported receiving an objection against a property valuation, claiming there was a risk of moisture rising from the ground. The AI model behind the objection falsely concluded that the property was near water and generated the moisture argument. Another objection cited noise and vibrations from a nearby mill, even though the mill had been gone for 50 years and there was no actual disruption.
Verloop explained the complexity of dealing with AI-generated objections. “AI models are often incredibly smart, but they can still be dumb sometimes,” he said. His municipality, along with at least nine others in the country, has begun using AI to process objections. “With the rising staff shortages, particularly in the Achterhoek region, it’s becoming nearly impossible to manage this manually,” Verloop added.
Montferland has developed its own chatbot to assist residents around the clock, even when government offices are closed. The bot is currently trained to provide general information but is expected to expand its capabilities, helping residents with everything from building permits to local taxes. Other cities, including Amsterdam, Breda, and Zaanstad, are also experimenting with AI for municipal services, while the Association of Dutch Municipalities (VNG) is conducting research into the role of AI in handling objections.
Marijn Bodelier, a lawyer with the international firm Greenberg Traurig, highlighted the growing problem of increased objections. “There is already a significant issue. The number of cases at the Council of State is rising sharply. People are objecting to everything, from housing to the energy transition. Now, with AI, it’s becoming incredibly easy to challenge permits. With AI, this issue is amplified a thousandfold. Municipalities are focused on how to handle existing objections, but they underestimate what is coming their way.”
Verloop agrees with this assessment, stating that many municipalities are just starting to grasp the scale of the problem. “Most municipalities are just beginning to wake up to it.”
Despite the challenges, experts agree that AI is not yet a perfect legal tool. Berke Aslan, a student in econometrics, co-founded Bezwaarfabriek, a company that developed an AI model for mass objections against traffic fines. Aslan acknowledged the limits of AI, stating that the model is still evolving. “What people often overlook is that the burden of proof for a traffic fine lies with the individual, and most people don’t have that evidence.”
Aslan’s company plans to scale its objections process and focus more on businesses rather than individual citizens due to high competition. Other companies, such as Appjection and Bonnetje, are also exploring AI-driven objection services.
Verloop noted that the need for such businesses may eventually diminish. “With AI, anyone can be their own lawyer or legal advisor,” he said.
