AI agents are flooding public services with new requests

Public services around the world are seeing a significant surge in applications and requests, with complaints to the housing ombudsman in the UK more than doubling since the introduction of ChatGPT. The US Consumer Financial Protection Bureau (CFPB) has also seen a 5x growth in complaints over the same period. Researchers are tracking this trend as part of a broader trend they call 'agentic flooding,' where AI tools are changing the way people interact with public services. Dr. Chris Schmitz, a researcher, has been studying this phenomenon and found that 84 different cases of potential flooding across 11 jurisdictions show broad evidence of this trend. Schmitz believes that AI is making it easier for people to interact with public services, as it can automate tasks such as filing forms and submitting claims. While some of the new filings are adversarial, many are from legitimate applicants using AI to file claims that would otherwise be abandoned. Schmitz sees this as an opportunity to remake social services for the AI era. He notes that most cases have not seen a slowdown in growth, suggesting it will likely continue for years to come. According to Schmitz, the increasing skills and availability of AI are driving this trend. He believes that public services can benefit from this trend, as it can help streamline processes and reduce administrative burden. Schmitz says that the vast majority of cases he has found are from people who are entitled to claim for something, claiming for that thing. The surge in applications and requests is similar to what many bug-bounty services experienced last year, as companies found their inboxes flooded with low-quality reports generated by LLMs. However, Schmitz notes that most of the new applications to public services are coming from real people with legitimate claims. Schmitz's research suggests that public services can use AI to improve efficiency and reduce costs. The researcher believes that this could be the moment to remake social services in a more AI-friendly way. Schmitz notes that making AI go well requires being able to detail out what the good version of things looks like. He says that this could be an opportunity to create a better version of services that are currently difficult to access. Schmitz's research has been published in a paper set to be presented next month at the AI Ethics and Society conference. The full dataset is hosted online for reference.