As generative AI reduces the barriers to filling out forms, writing complaints, and organizing documents, public service systems in many countries are experiencing a new wave of application peaks. TechCrunch reports that in recent years, there has been a noticeable increase in the number of applications and complaints from some institutions in the UK, the US, Brazil, and Germany.
Applications continue to rise in multiple regions.
Researcher Chris Schmitz refers to this phenomenon as "proxy peak." In a paper that will be submitted next month to the AI Ethics and Society conference, he reviewed 84 cases from 11 judicial jurisdictions and argues that AI tools are changing the way the public interacts with government service systems.
The report mentioned that the number of complaints received by the UK Housing Ombudsman has increased from 2,600 in 2022 to over 7,000 last year. During the same period, the number of complaints received by the US Consumer Financial Protection Agency increased by about five times. Similar increases were observed in judicial petitions in Brazil and parliamentary petitions in Germany.
AI Lower the submission threshold
This study does not directly conclude that AI is the sole reason for the surge in applications, but most cases show a similar trend: data remained relatively stable around 2022, and thereafter, with the widespread adoption of AI tools, the number of submissions began to rise steadily and continuously, and there is no sign of a significant slowdown yet.
Schmitz believes that the reason is not complicated. In the past, applicants often had to organize a large amount of background information themselves and repeatedly revise their wording. Now, users only need to paste a letter or take a photo of a document and submit it to tools like Claude or ChatGPT to quickly generate a response or appeal text that can be submitted.
Review pressure increases simultaneously.
Such changes also put public institutions under greater processing pressure. Reports mentioned that last year, many vulnerability bounty platforms were overwhelmed by low-quality reports generated by large models. Although most of these did not constitute serious security issues, the platforms still had to review them one by one, resulting in a significant increase in resource consumption.
However, Schmitz believes that public service scenarios are not entirely the same as vulnerability reporting. The latter often contains a large amount of invalid content, whereas a considerable portion of new applications for public services come from genuine applicants, and their demands are legitimate in themselves.
He pointed out that many people did not apply in the past not because they were not qualified, but because the process was too complicated and the cost of preparing the required documents was too high. Now, AI has reduced this administrative burden, which may allow legitimate applications that were previously abandoned to be re-introduced into the system.
Reconstructed or part of a common process
In his view, this is not only an operational pressure for the public sector but could also become an opportunity to restructure service processes. If AI is already able to assist users with tax filing, complaints, and welfare applications, then the public service systems may also need to be redesigned in terms of their entry points, review processes, and response methods to adapt to new usage habits.
