Security and governance lessons from internet protocols for ai agents
From TCP/IP to MCP: Security and Governance Lessons from Internet Protocols for Agentic AI
Computers and Society
Summary
New protocols help AI agents talk to people and tools, similar to how internet protocols connect computers. The paper finds that if security isn’t prioritized early in AI protocols, fixing problems later will be much harder and costlier. It also highlights that current control over these AI protocols is concentrated among a few companies, which could be risky if these systems become very important. The authors suggest ways to improve both security and governance for safer AI development.
What this means in practice
- •For ai protocol developers: Design AI agent interaction protocols that embed strong security measures from the start to avoid costly fixes later.
- •For technology policymakers: Create governance frameworks that distribute control of AI agent protocols to ensure legitimacy and prevent centralization risks.
A position paper. It proposes an approach and reports no results.
Authors
Ben Bucknall, Andrew Byrley, Lewis Hammond
Abstract
The past two years have seen considerable interest in protocols for managing how AI agents interact with human users, external tools, and each other. In many ways, these protocols, including the Model Context Protocol (MCP) and Agent2Agent (A2A), mirror the suite of protocols that underlie the modern internet. Motivated by this parallel, this paper draws lessons from the development, history, and governance of internet protocols with the aim of informing the responsible development of AI agent protocols. Based on our analysis, we make two conclusions. First, if AI agent protocols do not sufficiently prioritise security in the early stages of development, they risk accumulating a security deficit that becomes increasingly costly to address over time due to practical and economic challenges. Second, current governance structures for AI agent protocols centralise decision-making power in a small number of individuals and companies, raising important questions regarding legitimacy, especially if they become critical digital infrastructure. We conclude with a discussion and implications for policymakers and developers of AI agent protocols.