AINEWS
Citizens Are Filing 300-Page AI-Generated Complaints. Local Governments Are Drowning.
Four in five Danish municipalities report a rise in AI drafted complaints, and nearly half say the new filings cost more caseworker time than they save.
Four in five Danish municipalities report a rise in AI drafted complaints, and nearly half say the new filings cost more caseworker time than they save.
When a resident of Holstebro, Denmark, submitted a complaint topping 300 pages, deputy director Claus Brandt's team at Holstebro Municipality took three working weeks to respond. Two hours after the reply went out, a 100-page follow-up arrived. The complaint had not been written by hand. According to Ingeniøren's survey of Denmark's 98 municipalities, it was drafted with generative AI, the kind of citizen request now arriving routinely in Danish town halls.
This is a public-administration story about the citizen side of the counter, not a tech-industry story. Half of Denmark's municipalities responded to the survey, and the pattern across them is unusually consistent. Thirty-nine of roughly fifty reporting bodies said they had seen a rise in citizen inquiries or complaints believed to be wholly or partly AI-authored. Thirty-two flagged that the new complaints are longer and more legally complex than before. Twenty-four said the AI-generated filings are slowing case processing, a net resource cost rather than a gain. Twenty-seven noted potential upsides for citizens who previously could not navigate the paperwork.
The numbers are professional judgement rather than audited counts, since municipalities do not, in the ordinary course of business, tag incoming mail as "AI-drafted." But the pattern matches a stack of named cases the same Ingeniøren report walks through. In Haderslev, family-law cases now routinely cite multiple statutes simultaneously. In Favrskov, AI-cited laws can be factually incorrect. Silkeborg reports cases where the AI is directly misleading because it invents or quotes outdated legislation. Frederiksberg finds that AI complaints argue at a general level but miss the case-specific facts a caseworker actually needs.
Aalborg's children and young people department says AI currently consumes more resources than it frees up. Albertslund reports a quieter shift: non-Danish and non-English speakers are now filing legally complex, grammatically correct Danish complaints they could not have written on their own. Affected services span social and health, children and family, employment and labour market, and general citizen advisory, the part of municipal government that runs on letter, email and case file.
The mechanism is the part most public-sector AI strategy has yet to price in. Governments have spent two years debating how AI will speed up the inside of the bureaucracy, drafting memos, summarising case files, answering routine calls. Generative AI has lowered the cost of something else first: the complexity of the citizen side of the conversation. Anyone with a chatbot and a case can now produce a lengthy, multi-statute complaint in an afternoon. The bottleneck has moved.
The constructive question, and the one the survey leaves open, is what local administrations change next. The candidate moves are visible in the same responses: triage layers for length and complexity, AI-assisted summarisation tools for case workers rather than for citizens, clearer citizen-facing guidance on how to file a usable complaint, and staffing reallocation toward the family-law and benefits offices where the multi-statute cases now land. The Holstebro detail suggests the pressure is not theoretical: three working weeks for one reply, then a 100-page follow-up within two hours.
It comes as no surprise to three experts in digitalisation that Danes have embraced artificial intelligence—even when they are sending complaints to the local authorities. It is well known that language models can produce unnecessarily detailed, long, and complex complaints, explains Naja Holten Møller, associate professor at the Department of Computer Science at the University of Copenhagen. "It's a problem for the way we communicate with one another, because it leads to what is known as AI slop. This means that what we write often ends up being much longer and, in fact, often leaves the recipient unclear as to what the message actually is," she says. Per Andersen, professor of legal history and digitalisation of law at the University of Southern Denmark, also points out this challenge. According to him, general-purpose language models such as ChatGPT cannot be expected to have a handle on legislation and legal practice. "The answer you get from your language model often reinforces the idea that you're right, but that doesn't necessarily mean you are," he says, adding that there is a possibility of seeing a growing number of complaints that, from a strictly legal point of view, are unfounded. Ursula Plesner, associate professor of digitalisation in the public sector at Copenhagen Business School, emphasises that AI complaints can become a "very complex task" for the administration. "There's potential for democratisation in the fact that you don't have to be a lawyer to take up the fight against something within the system, but it puts a lot of pressure on case processing because of the increased workload," she says.
The demand side of public-sector AI is doing the opposite of what the strategy decks promised, and Danish town halls are the first to show what that looks like in practice.