AI floods public services with new requests

Researchers have identified a repeating pattern across 11 countries: since large language models became widely available, the volume of submissions to public services has multiplied — from housing complaints in the United Kingdom to parliamentary petitions in Germany.

The numbers behind the surge

In the UK, complaints to the Housing Ombudsman jumped from 2,600 in 2022 to more than 7,000 last year. In the United States, the Consumer Financial Protection Bureau recorded a fivefold increase over the same period. Similar spikes appear in judicial petitions in Brazil and petitions to the German parliament. Chris Schmitz, who will present the findings next month at the AI Ethics and Society conference, has mapped 84 such cases and labelled the phenomenon "agentic flooding."

Correlation, not proven causation, but consistent

The paper stops short of asserting direct causality — the methodology does not allow it — yet in most cases the curve is identical: flat submissions until 2022, then accelerating growth as AI tools spread. Perhaps more important, the rate has not yet begun to slow in most jurisdictions, suggesting the trend will continue for years. Schmitz explains that the technical barrier has collapsed: where users once needed precise prompts for ChatGPT 3.5 and manual context gathering, they can now photograph a letter in Claude and obtain a usable draft in a single step.

Not spam — real people with real rights

The obvious comparison is to bug-bounty programmes, which were deluged last year with low-quality reports generated by language models. There, most submissions were worthless, yet companies were forced to review each one. Schmitz stresses a fundamental difference: "The vast majority of cases we found are people who are entitled to claim something, claiming that thing." In public-policy terms, what blocked them before was "administrative burden," and AI is now removing it.

An opportunity to redesign the process

For Schmitz, the current wave is a rare chance to rethink service design itself — to make services friendly to intelligent agents instead of fighting them. "Anyone who has used a chatbot for a tax refund knows there is a good version of this, where you actually get help," he says. The task is large and has barely begun, but it points in an optimistic direction: instead of treating the surge in requests as a burden, it can be seen as evidence that rights are finally becoming accessible to people who previously gave up on them.