Why the Future of AI Depends on Protocols, Not Just Platforms
The rapid development of Artificial Intelligence requires a change in how regulators and industry leaders manage governance and market structure. Tim O’Reilly and the AI Disclosures Project examine the shift from static information disclosures to functional, protocol-based systems that shape the "agentic economy."
As Tim O’Reilly argues regarding the necessity of transparency: "You can’t regulate what you don’t understand."
In a digital economy, protocols are more than just code; they are "standardised ways of doing something" that facilitate large-scale coordination. They represent society's communication and control systems. Whether it is GAAP (Generally Accepted Accounting Principles) facilitating trust between investors and firms, or road signs dictating the flow of traffic, protocols are the invisible infrastructure of progress.
The transition to the agentic economy is currently in its "path routing" phase, an early, experimental stage in which many approaches will fail. Just as the internet evolved from fragile path routing to scalable domain routing, the AI economy must develop towards a decentralised, scalable infrastructure. The choice is straightforward. We can accept a future of "engineered agreements" where a few gatekeepers claim the value of the AI revolution, or we can advocate for a future of "engineered arguments", featuring modular, decentralised architectures that enable experimentation with business models, payment systems, and quality signals.
In other words, will we build an AI economy dictated by a few tech giants, or will we develop the mechanisms that enable a truly open, human-agent market to thrive? The answer depends on the protocols we choose to develop today.
To build a functional agentic economy, we must design the rules and incentives that guide self-interested actors toward outcomes that benefit the whole. The current "war" between AI labs and content owners over IP is a failure of mechanism design that requires immediate intervention, according to O'Reilly and the AI Disclosures Project.
