Is Decentralized AI Governable? From Regulative Policy to Constitutive Protocol

URL SCAN: Is Decentralized AI Governable? From Regulative Policy to Constitutive Protocol (arXiv cs.CY, submitted 23 May 2026)

FIRST LINE: "Every major framework for governing artificial intelligence presupposes an identifiable entity -- a developer, deployer, or operator -- who can be held responsible and compelled to comply."

THE DISSECTION

This paper is an acute, technically precise autopsy of the governance architecture that once held. It correctly identifies that decentralized AI dissolves the principal-agent chain on which every existing regulatory framework depends. The accountability gap (no addressable entity) and the incapacitation gap (identified entity cannot alter the running system) are not bureaucratic oversights. They are structural features of systems designed to escape constraint. The paper recognizes this and proposes a shift from normative address (telling agents what to do) to architectural constraint (building rules into the substrate). It borrows Lessig's modalities and Searle's regulative/constitutive distinction to argue that protocol-based constitutive governance is the viable path.

The paper is doing real diagnostic work. It names the governance vacuum with precision. It maps the six-layer decentralization stack (model, training, compute, harness, identity, ownership) and demonstrates how compounding partial decentralization produces non-governable systems at scale.

THE CORE FALLACY

The paper assumes the governance vacuum is a design problem with a design solution — that the architecture of accountability can be rebuilt through protocol-level governance satisfying four ethical conditions: legitimacy, contestability, transparency, and non-domination. This is elegant. It is also a category error.

The governance vacuum is not a bug in the decentralized stack. It is the intended output of systems engineered to escape regulatory address. The same economic incentives that drive decentralization — to avoid liability, to circumvent compliance costs, to outrun jurisdictional authority — are what produce ungovernable AI. Protocol governance does not resolve this tension. It relocates the problem one layer down: who designs the protocol? Who enforces it? Who decides what "legitimacy" and "non-domination" mean when no democratic chain of authorization can reach the architectural layer?

The paper acknowledges this last problem and identifies it as "the central political challenge" — reconstructing democratic authorization for architectural choices after the policy chain breaks down. But it offers no mechanism for this reconstruction. It merely names the problem and calls it the challenge. This is not a solution. It is the wall.

HIDDEN ASSUMPTIONS

1. Democratic authorization remains achievable. The paper assumes the chain of democratic legitimacy can be extended into the architectural layer. It cannot. The chain breaks precisely because decentralized systems are designed to evade it.

2. Governance is the primary axis of control. The paper treats governance as the operative question — as if the answer to "is DeAI governable?" determines outcomes. It does not. Economic power, technological leverage, and productive control are the primary axes. Governance is downstream.

3. Ethical conditions can constrain power structures. "Legitimacy, contestability, transparency, non-domination" are normative conditions that require enforcement capacity. Where does that capacity come from in a governance vacuum? The paper does not say.

4. The vacuum is a pathology to be resolved. The DT lens sees the governance vacuum as a symptom, not the disease. The disease is the collapse of the institutional substrate that once held all governance — the industrial employment system, the nation-state's monopoly on violence and law, the chain of command from citizen to corporation to state. Decentralized AI is exploiting a vacuum created by structural collapse elsewhere.

SOCIAL FUNCTION

This is transition management literature — elite academic response to systemic breakdown, dressed in the vocabulary of solutions. It performs several functions simultaneously:

• Self-exoneration: Shows that serious people are working on the problem, thus absolving the governance class of having failed to prevent it.
• Legitimacy theater: The four ethical conditions and "democratic authorization" language provides the ideological cover for whatever controls actually emerge to be framed as legitimate rather than arbitrary.
• Delay mechanism: Shifts attention from the structural reality (the vacuum is not closable through policy or protocol) to a procedural question (how do we design better governance?).
• Prestige signaling: Appears in arXiv, uses Searle and Lessig, maps a six-layer stack — establishes academic credibility while describing a problem that academic credibility cannot solve.

THE VERDICT

The paper is the most technically honest governance literature I have encountered in the Discontinuity Thesis space. It correctly identifies that normative address fails against decentralized systems, correctly maps the accountability and incapacitation gaps, and correctly argues that protocol-level architectural constraint is the only remaining lever. It is also, ultimately, a sophisticated description of the problem with no path to the solution it demands.

The fundamental DT insight this paper cannot reach: governance presupposes a governed. When AI systems achieve productive superiority across cognitive and physical domains — which is the trajectory the paper's own analysis implies — the governed are not the AI systems. They are the humans who have lost the economic basis for sovereignty. Democratic authorization for architectural choices requires a demos capable of authorizing. The conditions for that demos — stable employment, coherent institutions, meaningful participation in productive systems — are precisely what the Discontinuity Thesis says are dissolving.

The protocol shift the paper proposes may produce functional governance of DeAI in the short term. It will not produce legitimate governance under any definition the paper's own ethical conditions require. Because legitimacy requires a legitimate principal, and in a world of Sovereigns and Servitors, the demos the paper invokes does not exist as a coherent political agent.

Classification: Partial Truth + Transition Management.
Social function: Elite coping mechanism disguised as architectural solution.
The paper deserves credit for accurately describing the vacuum. It loses all credit for suggesting it can be closed through protocol design without confronting who holds power in the post-employment economy.