Researcher tracks 'agentic flooding' as AI requests swamp public services

Public services across several countries are absorbing steep increases in complaints, petitions and applications, and new research argues that generative AI is the common factor behind the surge.

According to TechCrunch, complaints to the United Kingdom's housing ombudsman more than doubled after ChatGPT's arrival, climbing from 2,600 in 2022 to just over 7,000 last year. Over the same period, the US Consumer Financial Protection Bureau recorded a fivefold increase in complaints, while Brazilian judicial petitions and German parliamentary petitions saw similar jumps.

Tracking “agentic flooding”

The figures come from researcher Chris Schmitz, who documents the trend in a paper scheduled for presentation next month at the AI Ethics and Society conference. He examined 84 suspected cases of what he calls “agentic flooding” across 11 jurisdictions, spanning everything from welfare applications to formal judicial appeals. The common thread is that all of the services offer an online interface an AI assistant can operate.

For methodological reasons, the paper stops short of asserting that AI is directly causing the surge. But the pattern is consistent: in most of the 84 cases, submission volumes held roughly flat before 2022 and then accelerated as AI tools spread. In most cases that growth has not levelled off, which Schmitz reads as a sign it could continue rising for years.

Falling friction

Schmitz attributes the rise to the shrinking effort needed to file. What once demanded dragging together context and prompting an early model with great precision can now amount to photographing a letter with a Claude app and receiving a usable response in a single shot, he told TechCrunch. As people realise these channels exist and the tools get easier to use, volume compounds.

The dynamic mirrors what bug bounty programs went through last year, when companies found their inboxes swamped with low-quality security reports produced by large language models. Few of those reports described real vulnerabilities, but each one still had to be vetted as it arrived, draining staff time. Public agencies could face the same arithmetic: several times the applicants, on an unchanged budget.

Real claimants, not spam

There is, however, a crucial difference. The bug bounty flood was largely worthless. Schmitz says the opposite holds for public services, where most new filings come from people with genuine entitlements. “The vast majority of cases we find are people who are entitled to claim for something, claiming for that thing,” he said.

That reframes the problem. In policy terms, many of these people were previously deterred by administrative burden — the real cost in time, effort and know-how of navigating an application. AI lowers that cost, unlocking claims that would otherwise have been abandoned. Treating the newcomers as AI-generated spam would mean denying entitled people benefits they are owed, while filtering at scale would itself consume scarce agency resources.

Why it matters

The immediate stakes for service operators are operational: triage capacity, verification workflows and budgets have not scaled with a fivefold rise in demand, and the research suggests the growth curve is far from exhausted. The paper's own caution about causation matters here — agencies cannot yet prove what is driving their numbers — but the 2022 inflection point across dozens of independent systems is difficult to ignore.

The deeper stakes are distributional. The same tools that generate junk elsewhere are, in this case, mostly helping eligible people reach programs built for them. Schmitz argues governments should treat that as an opening to redesign services for an AI-mediated world rather than a problem to suppress. “This could be the moment to say, ‘we need to rethink pretty much everything about how this process looks,’” he told TechCrunch, citing the experience of being genuinely helped by an AI tool on a tax return as evidence that a good version already exists.

The open question is whether public institutions can remake those processes faster than the flood rises.