AI Agents Need an Authority Layer for the Internet

The shift from users to agents

Policymakers continue treating artificial intelligence as just another software category—comparable to social media or cloud computing. But the proliferation of AI agents represents a more fundamental shift: not what the Internet does, but who acts through it.

The naive view treats AI agents as simply another type of user, digitally indistinguishable from a person browsing or shopping online. Yet the term "agent" reveals the deeper dynamic at play. These systems function as intermediaries between humans or organizations and the network itself, operating with delegated authority to navigate complex digital environments in ways their operators did not explicitly script.

Social scientists have long studied principal-agent problems in human contexts—elected officials, insurance brokers, real estate agents. These actors operate with incomplete information and may inadvertently act in misaligned ways. Agentic AI should be viewed through the same lens, but at massive scale.

The governance gap

Multiply these capabilities and asymmetries by millions of deployments, and the implications expand dramatically. Internet traffic will increasingly represent software acting on behalf of humans, often interacting with other software representing other humans and organizations.

The fundamental questions change. Authentication is no longer simply about verifying a user or device. The critical issues become: Which agent is acting? On whose behalf? Who authorized it? What exactly was it authorized to do? For how long? Under what conditions? Can those permissions be verified, constrained, audited, and revoked?

Today's policy debates remain siloed. Internet governance focuses on infrastructure, standards, interoperability, and digital rights. AI governance concentrates on models—how they are trained, whether they are safe, what outputs they produce. Neither conversation alone addresses the challenge of authority and accountability when human intent is increasingly executed through autonomous software.

Why it matters

The architecture being built today will determine whether human authority survives in an agentic Internet. If delegation chains become opaque, attribution may exist technically while accountability disappears in practice. The question is whether powerful platforms will control these relationships, or whether open standards will preserve user choice and interoperability.

Emerging standards and the authority layer

The standards landscape is not entirely empty. The Model Context Protocol is developing mechanisms for agents to access external tools and data, including an authorization framework based on OAuth. Google's Agent2Agent protocol, now hosted under the Linux Foundation, aims to allow agents from different vendors to discover one another and coordinate. The IETF is seeing proposals for agent identity and verifiable delegation.

These separate efforts are beginning to construct something more consequential: an authority layer for the Internet. Identity, delegation, authorization, and agent-to-agent interoperability will determine not just how machines communicate, but how authority moves between humans and machines.

The opportunity is to connect these standards around a common principle of verifiable human delegation. The alternative is an agentic Internet that emerges as a collection of interoperable protocols with no shared model of accountability.

Technical standards, while difficult, are achievable. The harder question is procedural: where does human authority sit when actions pass through chains of agents, platforms, and tools? Who decides how responsibility is divided when the negotiating actors are powerful companies?

Preserving human agency in an agentic Internet means designing systems where delegation is visible, permissions are meaningful, actions are attributable, and authority can be withdrawn. It means maintaining interoperability so users are not locked into a single agent or platform. And it means ensuring that when an agent acts, there remains a clear line back to the human or organization that authorized it.

The Internet became a global public resource because its architecture allowed different networks and communities to interoperate while leaving room for human choice. The future may be increasingly agentic, but it should remain human-directed, human-accountable, and human-centered.

These observations were originally published in Tech Policy Press.