The phrase "agentic flooding" sounds immediately negative, and I'm not convinced all of it is. The paper looks at what happens when AI dramatically lowers the work required for ordinary people to interact with government—applying for a benefit, appealing a denial, filing a complaint, requesting records, navigating a court procedure. A lot of these systems are difficult enough that understanding what you are entitled to and writing the required response can take hours. An LLM makes a significant portion of that work cheap. People submit more requests, and the requests become more complex.

The authors collected 84 potential cases across 11 jurisdictions. They are careful: the dataset does not prove AI caused every increase. But there is enough evidence that something real is happening, and we aren't even talking about highly autonomous agents yet. In 87% of cases, the enabling technology was essentially an LLM helping someone produce sophisticated text. Quantitative flooding (more submissions) appeared in 60% of cases; qualitative flooding (more complicated submissions) in 90%. Half had both.

Some examples are remarkable. German social courts reportedly attributed a 55% year-over-year increase in cases during 2025 partly to AI-generated claims. Australian officials have discussed bringing back fees for Freedom of Information requests. In one U.S. federal-court case the researchers studied, letters from self-represented litigants had grown to more than 4,000 pages.

There is room for abuse—terrible legal arguments, fraudulent claims, deliberate overwhelm. But most cases in the dataset are not described as adversarial. They are ordinary people using AI to interact with government, and that distinction matters.

The most interesting sentence in the paper for me is essentially that some of the most vulnerable government services have historically been protected by friction rather than design. A person may be legally entitled to appeal, but appealing requires understanding a complicated rule, finding the correct form, gathering evidence, writing a coherent argument, and meeting a deadline. Some percentage simply stop. Difficulty became part of capacity planning.

AI changes the economics. If something that took four hours now takes 20 minutes, more people do it. That looks like "flooding" from inside the agency. From the citizen's side, it might look like finally exercising a right they already had. A legitimate appeal does not become illegitimate because software made it easier to write. If making the process easier overwhelms the service, perhaps the service was depending on people failing to navigate it.

Governments can increase capacity, or they can suppress demand. Suppressing demand is much easier—charge a fee, add identity checks, rate limits, in-person requirements, block automation. The authors found governments had already introduced friction in 17% of the cases. That might solve the immediate capacity problem. It also recreates exactly the administrative burden AI was removing, and the people most affected are likely those who already had the hardest time. Someone with money pays the fee. Someone with a flexible job attends the appointment. Everyone else gets the CAPTCHA.

The harder response is redesigning services for the demand that actually exists—structured data, clearer eligibility, better APIs, automated routine processing, and AI assisting intake the way it already assists the public. In 25% of cases, agencies introduced AI tools in response. Eventually both sides of an administrative process may have software doing most of the mechanical work. At that point, perhaps we should reconsider why the process requires thousands of words passing back and forth in the first place.

Most of what they found isn't really autonomous agents yet—it is people using LLMs as assistants. Imagine when an agent can monitor every program someone may qualify for, prepare applications, track responses, draft appeals, and keep going until the process finishes. I think that could be incredibly valuable. Government services are supposed to exist for the public. If someone is legally entitled to something, making the process confusing enough that they never claim it should not count as successful scalability.

The danger is that governments rebuild the friction faster than the capacity. I hope we do the opposite.

Read "Characterizing Agentic Flooding of Government Services."

All notes