Decentralization risk starts with change control, not branding

Decentralization is risky when people confuse distribution with power. A token can be widely held, voting can be public, and the interface can look community-driven, while the actual power to change parameters, upgrade contracts, pause functions, or ignore execution still sits with a narrow operator set. IOSCO’s DeFi report is useful here because it treats decentralization as a spectrum across ownership, voting power, control of user assets, network design, and even off-chain infrastructure, and it explicitly warns that many systems retain hidden centralized authority despite decentralized branding.

For a token economy, that distinction is not cosmetic. Governance power decides emissions, treasury spend, whitelist policy, collateral parameters, oracle dependencies, fee switches, and upgrade paths. Those are not community vibes. They are levers. If a small actor set can move those levers quickly while the broader token base can only ratify or react, the protocol is structurally more centralized than its narrative suggests. IOSCO’s framing also matters because it refuses the lazy binary. The right question is not whether a protocol is decentralized. The right question is which layer is decentralized, which layer is not, and which layer actually governs outcomes.

The biggest risk of decentralization discourse is that it can hide the constitution. Teams often advertise token dispersion, community calls, or Snapshot participation. None of that tells you who can ship code, block withdrawals, coordinate emergency actions, appoint signers, or reject an approved proposal once lawyers enter the room. The power map matters more than the rhetoric behind decentralization claims.

Token voting does not reliably produce distributed power

Token-weighted governance usually starts as capital-weighted governance. A study of DAO voting power found that across several major DAOs, less than 1% of holders had 90% of the voting power. That is not a fringe edge case. It is a structural warning that “open participation” can coexist with extreme political concentration.

The academic record on governance tokens points in the same direction. A 2024 study of Compound, Uniswap, and ENS found that the majority of voting power was concentrated in a small number of addresses. Importantly, the authors also found that these large actors rarely overturned the outcome preferred by the wider community. That is a material nuance. Concentrated power is often latent rather than constantly abused. But latent power still matters because it defines who could block, accelerate, or shape outcomes when incentives diverge.

A broader 2023 study found that contributors held the necessary majority to control governance decisions in at least 7.54% of DAOs on average, and that contributors singularly decided at least one proposal in 20.41% of DAOs examined. The same paper observed ownership shifts before polls in 1,202 out of 8,116 evaluated proposals, or 14.81%. That does not prove manipulation in every case. It does show that voting power is mobile, strategic, and often concentrated around a relatively small political core.

Low participation deepens the problem. A 2025 DAO sustainability paper identified limited user participation, concentrated voting power, and technical design constraints as recurring governance weaknesses, while a 2026 follow-up using data from more than 50 active DAOs, 6,930 proposals, and 317,317 unique voting addresses reported persistently low participation and concentration of proposal activity. The point is simple: a token economy does not become politically plural just because many wallets exist. It becomes plural when many independent actors can meaningfully set agendas, veto harmful moves, and sustain participation over time.

The positive evidence is also clear. A 2024 large-scale analysis of 100 DAOs found that higher grassroots participation correlated with higher decentralization, and lower variance in voting power correlated with higher decentralization as measured by Gini metrics. Decentralization is therefore not a myth. It is a measurable governance outcome. It is just much rarer than the label implies.

The exception layer is the real constitution

Emergency powers reveal who actually governs a protocol. Ordinary voting tells you how a system behaves in calm periods. Exception mechanisms tell you who rules when markets gap, contracts break, or legal risk appears. That is why the most important governance question is often not “Who can vote?” but “Who can act without waiting for a full vote?”

Protocol Who can change or block outcomes Key parameter Why it matters
Compound v2 COMP delegates can propose and pass upgrades, while a Pause Guardian held by the Community Multi-Sig can disable selected functions. 25,000 COMP to propose, 400,000 votes needed in favor, and at least about 1 week to execution. Agenda access is gated, and emergency braking power is concentrated even inside token governance.
Uniswap An on-chain governance proposal requires a delegate with substantial delegated voting power. 1M UNI delegated to submit an on-chain proposal, 40M UNI voting in favor to pass, and a 10-day vote. Formal openness coexists with a high barrier to agenda-setting.
Maker MKR voters can trigger Emergency Shutdown through the Emergency Shutdown Module or by Executive Vote. 50,000 MKR deposited into the ESM can trigger shutdown. There is a last-resort kill switch because fully removing emergency power would increase system fragility.
Optimism A Security Council is designed to hold admin keys and can act without direct governance approval in defined emergencies. The council operates a Gnosis Safe with a 75% threshold. Security improves versus a single operator, but concentrated emergency authority remains explicit.
Arbitrum A 12-member Security Council can take emergency and non-emergency actions, and emergency meetings can be convened without prior notice and without a quorum requirement for the Foundation Directors. A majority of Foundation Directors may reject an approved AIP on specified legal, contractual, or fiduciary grounds. Token voting is not the whole constitution. The legal wrapper has real veto surface.
Lido V3 stETH holders can use Dual Governance to veto LDO governance decisions through Veto Signalling and Rage Quit. Veto Signalling begins at 1% of total stETH supply, can block governance motions for 5 to 45 days, and Rage Quit withdrawals face an additional 60 to 180 day timelock. This is a rare example where the exception layer distributes counter-power to users exposed to governance risk.

The table shows why broad claims of decentralization are weak unless they describe exception rights. Compound and Uniswap make proposal power expensive. Optimism and Arbitrum formalize emergency councils. Maker retains a shutdown path. None of these choices are irrational. Each exists because governance needs operational flexibility under stress. But every one of them also defines a smaller group whose decisions matter more than the median token holder’s vote.

Lido is the useful counterexample. Its Dual Governance design does not simply add another committee. It gives stETH holders an explicit veto channel against LDO governance and a path to exit through Rage Quit. That is a genuine redistribution of power from governors to governed users. Evidence like this matters because it cuts against the lazy assumption that every extra governance layer is centralizing. Some layers are checks, not capture.

Delegation solves turnout and creates a political class

Delegation is usually sold as a decentralization aid. In practice, it often becomes an elite formation mechanism. The 2025 paper on token delegation found that DAO delegation systems frequently reinforce visibility bias, allowing a small number of highly ranked delegates to accumulate disproportionate influence, and that those delegations are often misaligned with token holders’ expressed priorities. That is the classic political trade-off. Delegation lowers participation costs, but it also manufactures a durable class of intermediaries.

This does not mean delegation should be removed. Without it, turnout often collapses and governance becomes even easier for insiders to dominate. The problem is that many DAOs stop the analysis at “delegation exists,” when the harder questions are who receives it, how sticky it is, how discoverability is ranked, whether delegate competition is real, and whether token holders can monitor drift. A governance system with 100,000 holders and 15 effective delegates is not broadly decentralized in any politically meaningful sense.

The deeper risk is temporal. Governance concentration is not just about who holds tokens on average. It is also about who can assemble power at the right moment. Kitzler and coauthors’ finding that token ownership shifted shortly before 14.81% of observed proposals suggests that governance power is not static. It can move around events, incentives, and controversies. That makes delegation systems and vote snapshots part of the attack surface, not just the user experience layer.

Pure on-chain governance breaks at the legal and operational edge

Naive decentralization struggles when a protocol has to sign contracts, hire contributors, hold IP, or answer to courts and regulators. That is why many major systems converge toward hybrid governance structures. Arbitrum’s bylaws are unusually explicit. They state that the Foundation represents tokenholders in contractual and legal processes, that a majority of Foundation Directors may reject an approved AIP on specified legal, contractual, or fiduciary grounds, and that Foundation Directors are not fiduciaries for tokenholders. In power terms, that means the legal shell is not a neutral wrapper around token voting. It is an institution with its own decision rights.

Uniswap’s move in the opposite direction makes the same point. In August 2025, the Uniswap Foundation proposed DUNI, a Wyoming DUNA, to give Uniswap Governance legal recognition, contracting ability, treasury management capacity, and participant liability protection. The association agreement then limits administrators and agents to authority granted through a validly executed governance proposal. That is an attempt to translate on-chain governance into a legally legible structure without giving administrators open-ended discretion.

This is the right way to read hybrid governance. It is not evidence that decentralization failed. It is evidence that pure token voting is often too thin to run real organizations. The risk is different. Once named administrators, foundations, councils, or agents are necessary for execution, governance analysis has to include jurisdiction, liability, authority scope, replacement procedures, and conflict resolution. Otherwise the protocol can look decentralized onchain while remaining highly legible to capture offchain. IOSCO explicitly includes off-chain infrastructure and control of enterprise functions in the decentralization analysis for that reason.

What serious governance diligence should measure

The correct response to decentralization risk is not to demand “more community” in the abstract. It is to map power with enough precision that trade-offs become visible. For token economy work, that means at least six questions.

At FinDaS Tokenomics, this is usually where tokenomics consulting becomes governance analysis. Before debating emissions, utility, or incentives, the first pass should map parameter control, veto rights, emergency keys, delegated blocs, legal signatories, and execution dependencies. A token economy is only as decentralized as the smallest group that can still force the outcome that matters.