AI Agents Are Flooding Public Services With New Requests

Government benefit systems designed for human applicants are being hit by a wave of agent-submitted claims — and there's a twist in the data

Published: 2026-09-11 Category: Quick Take Sources: TechCrunch

The agent wave hits government

Public services are becoming the latest front line of the AI agent boom. Researchers are observing a surge in requests submitted to public benefit systems by automated agents, rather than by humans sitting at keyboards. The systems were built to assume a human applicant in the loop — a person who filled out a form, made a mistake, and might call to correct it. Agents change that assumption entirely.

The surprising finding, though, is that the flood is largely legitimate. As one researcher told TechCrunch: "The vast majority of cases we find are people who are entitled to claim for something, claiming for that thing." In other words, agents are not primarily being used to commit fraud by gaming systems; they are being used to help eligible people actually claim benefits they were previously too encumbered to pursue.

Two readings of the same data

That is a genuinely interesting split. The cynical read is that anyone can frame the data charitably, and that agents will inevitably be weaponized for fraud at scale even if they are not yet. The generous read — which the researcher's framing supports — is that agents are democratizing access: the paperwork barrier that once kept eligible claimants away is dissolving because an agent can fill out the forms, gather the documentation, and escalate correctly.

Both things may be true at once. Agents lower the activation energy for legitimate claims, which is a feature. They also lower the activation energy for abuse, which is a risk. The systems that serve the public now need to distinguish between those while operating at volumes they were never designed to handle.

The infrastructure gap

The deeper lesson is that agent adoption is outrunning institutional capacity. Public services are built around human interaction, identity verification, and appeals — all friction that agents can often bypass or scale past. Agencies now face a choice: treat the agent wave as a hostile intrusion to be locked down, or as a signal that the interface to government should be rebuilt for programmatic access. The most likely outcome is a messy hybrid, but the direction of travel is clear: when agents are involved, "everyone is a claimant, and everyone is a potential abuser," and institutions have to be designed for both realities at once.

Source: TechCrunch, "AI agents are flooding public services with new requests" (Sept 10, 2026).