Glostrup Municipality admits using AI in case handling

Glostrup Municipality has acknowledged using AI tools to process a resident’s case, violating its own guidelines, Danish broadcaster DR reports.

Jonas Helmer, who has muscular dystrophy, suspected AI involvement after receiving an 18-page decision from the municipality. A specific formulation in the document struck him as uncharacteristic of human writing. After requesting access to case files, he confirmed his suspicion: a municipal employee had used AI—specifically Copilot and ChatGPT—to assist in handling his request for disability-related expense coverage.

The municipality admitted in an email to Helmer that this was a “clear error.” Its internal rules permit AI use but explicitly prohibit its application in case processing, decision-making, or assessments.

Helmer remains skeptical, particularly about whether his personal data was input into the AI tools. The municipality denies this but cannot verify the claim, as the employee deleted their chat history and prompts. “I don’t trust a word of it. They say they didn’t upload sensitive personal data, but there’s no way to check,” Helmer said.

Allan Frank, an IT security specialist at the Danish Data Protection Agency, told DR that if personal data was indeed processed in AI tools, it would likely violate data protection rules. He noted that the agency has received at least 50 similar reports in recent years.

Frank highlighted multiple potential legal breaches, including violations of GDPR’s transparency principle and the municipality’s obligation to account for data processing. Additionally, sharing personal data with private AI providers without legal basis could constitute another violation, as public authorities in Denmark generally lack such authorization.

Glostrup Municipality declined an interview but stated in an email that it is now clarifying its AI usage guidelines to ensure compliance across the organization.