Dutch AI-powered parking scanners issue hundreds of thousands of wrongful fines
Dutch municipalities deployed AI-powered parking scanners that the data protection authority found issue between 500,000 and 750,000 wrongful fines annually. The systems fail to recognize legal parking exceptions like loading zones and disability permits, and misidentify license plates, charging wrong vehicle owners.
What happened
Dutch municipalities deploy AI-powered parking scanner vehicles ('scanauto's') that use artificial intelligence and algorithms to automatically scan vehicle number plates and issue parking fines. The Autoriteit Persoonsgegevens (AP), the Dutch data protection authority, conducted an investigation and found that the scanners frequently issue incorrect fines.
The scanners cannot distinguish between vehicles that are temporarily stopped for loading or unloading (which is permitted) and illegally parked vehicles. They also fail to detect disability parking permits displayed in vehicles. Additionally, the optical character recognition system sometimes misreads license plate characters and numbers, issuing fines to the wrong vehicle owners, including cases where incorrect fines were issued for over two years.
Reported impact
The AP estimates that more than 10 percent of fines issued are incorrect, resulting in approximately 500,000 to 750,000 wrongful fines annually across an estimated 40 municipalities using the scanners. The scanners conduct between 250 million and 375 million checks yearly, resulting in 3 to 5 million total fines.
Citizens must lodge formal objections to challenge fines, a process described as time-consuming and unclear. People are often required to pay fines while their objections are being processed, which can cause financial hardship, particularly in cases of repeated errors. In municipalities studied, between 300,000 and 450,000 appeals succeed annually, indicating a systemic problem.
Index analysis
The failures stem from the scanners' inability to assess contextual circumstances that human parking enforcement officers would recognize, such as loading zones or disability permits. The systems were designed for speed and scale rather than accuracy in determining whether fines are justified.
Governance controls
The AP advised municipalities to deploy more human parking enforcement officers to conduct on-site inspections and assess mitigating circumstances. In the municipalities investigated, only two regularly conducted on-site verification of whether fines were justified. Human reviewers use additional contextual information including vehicle history, color, and model, as well as photographs of the surrounding area taken by scanners, to determine if fines are appropriate, but this review process does not resolve the initial detection errors.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.