Good governance gives token holders binding control over a written list of decisions and makes every change wait long enough to be checked. Arbitrum’s constitution, for example, sets a minimum of about 27 days from temperature check to execution for ordinary proposals and about 42 days for constitutional ones. A 12-member security council can act at once in an emergency, but only with 9 signatures.
Who decides this
The foundation council owns the governance design and is accountable for it working as written. That means deciding what the token vote can bind, how fast it can act, who can act in an emergency, and what record proves each decision. Counsel drafts the documents, and our DAO legal structure guide, linked below, covers the wrapper itself. The council owns the operating rules inside it.
The stakes are legal as well as practical. In June 2023 a US court, in a default judgment, held that Ooki DAO was a “person” under the Commodity Exchange Act and imposed a $643,542 civil penalty in a CFTC action. Governance design is part of how a protocol shows who was, and was not, in control.
What should the token vote control, and what stays with the council?
The token vote should control a defined list of decisions, and the foundation should implement passed votes within that list without re-deciding them. Anything outside the list stays with the council or needs a constitutional change.
Uniswap’s Wyoming association, DUNI, shows one published version. Its association agreement lets passed proposals change smart-contract parameters, appoint administrators, amend the governing principles, approve asset disposals and authorise pay for services. Its ministerial agent may only perform non-discretionary tasks, such as paying approved funds and keeping records. It cannot make policy, and members can widen or revoke its authority by vote.
Scope also has to cover the foundation itself. In December 2025 Uniswap governance passed the “UNIfication” proposal with about 99.9% support and more than 125 million UNI in favour. It switched on a protocol fee used to burn UNI and moved the foundation’s teams and responsibilities to Uniswap Labs. A token vote reorganised the foundation’s own role.
Our view: write a scope list before launch, in three columns. Decisions the vote binds. Decisions the vote may propose but the council may decline, on stated grounds such as illegality, sanctions or insolvency. Decisions that need a constitutional vote, including any change to the foundation’s own mandate.
How high should the thresholds be, and how long should changes wait?
Set thresholds high enough that a small coalition cannot pass changes alone, and delays long enough for holders to exit or object before execution. Each lever buys safety with speed.
Arbitrum publishes every number in its DAO constitution. A proposer needs at least 1,000,000 delegated votes to submit, or 500,000 for a temperature check. Voting runs 14 days; if quorum is reached in the last 2 days, voting ends 2 days after quorum. Quorum is measured against delegated votable tokens: 50% for constitutional proposals, bounded between 150 million and 450 million tokens, and 40% for ordinary ones, bounded between 100 million and 300 million.
| Lever | What it protects against | What it costs | Arbitrum’s setting |
|---|---|---|---|
| Proposal threshold | Spam and hostile proposals | Small holders must go through a delegate | 1,000,000 delegated votes |
| Quorum | A low-turnout vote passing a major change | Proposals fail when delegates are inactive | 40% or 50% of delegated votable tokens, with floors and caps |
| Voting period | Votes timed for when holders are not watching | Slower decisions | 14 days, extended if quorum arrives late |
| Waiting periods after passage | Executing a harmful change before anyone can react | Weeks before a fix takes effect | Fixed waits on L2 and L1, with a rollup challenge period between |
| Two classes of proposal | Changing the rules as easily as the parameters | More process to design and explain | Constitutional and non-constitutional, with different quorums |
The floor and cap matter more than they look. A pure percentage can become unreachable if delegation falls, or trivial if it is concentrated. Bounds keep quorum tied to real token numbers.
Our view: a new foundation should start with a constitutional class for changes to scope, treasury limits and emergency powers, and a lighter class for parameters. Review the numbers each year against actual turnout.
Should holders vote directly or through delegates?
Few holders can follow every proposal, so a delegate system is the realistic design, but the council must know who the deciding delegates are. Delegation concentrates power by design.
A European Central Bank staff working paper, reported by Cointelegraph in March 2026, studied Aave, MakerDAO, Ampleforth and Uniswap using 2022 and 2023 snapshots. In each, the top 100 holders held more than 80% of supply. The top 18 Uniswap voters held 52% of delegated votes, and the top 10 MakerDAO voters held 66%. About a third of top voters could not be publicly identified.
The paper states that it reflects its authors’ views, not ECB policy. The Uniswap Foundation responded that the data was old, that the DAO treasury is among the largest holders, and that excluding exchange cold wallets brings concentration to no more than 43%. Both readings support the same design point: concentration must be measured, not assumed.
Our view: publish a delegate register with each delegate’s statement, affiliations and conflicts. Track voting power held by the foundation, the labs company, investors and exchanges separately. Decide in advance whether treasury or team tokens may vote at all. That last decision has legal weight. In a US case, Samuels v. Lido DAO, a November 2024 ruling let the plaintiff’s claim proceed. The plaintiff alleges the DAO is a general partnership and that large early investors active in governance are partners. That was a ruling on a motion to dismiss, not a finding, and the allegations remain unproven.
Who can act in an emergency, and who watches them?
A small, elected council should hold emergency power, with a high signing threshold, a fixed term and a duty to report every use. Without one, a live exploit waits for a 14-day vote.
Arbitrum’s security council has 12 members who sign a multisig. Emergency actions need 9 of 12 signatures and take effect without delay. Non-emergency actions need the same 9 of 12 and then pass through the normal waiting periods. Elections run every 12 months, each replacing a cohort of six. Each election takes about 49 days, including a 14-day compliance review.
The trade-off is plain. An emergency body that can act instantly can also act wrongly instantly. The checks are the threshold, staggered terms and a rule that every emergency action is explained publicly and ratified or reversed by the token vote.
Our view: define “emergency” narrowly, in writing, before launch. Limit emergency powers to pausing and patching. Treasury movement should never be an emergency power. Treasury controls are covered in our treasury governance guide, linked below.
How does governance oversee money it hands out?
Every grant or programme the vote funds should carry milestones, a reporting schedule and a recall path, written into the proposal before the vote. Adding them afterwards takes another fight.
Arbitrum’s Gaming Catalyst Program shows why. In June 2024 the DAO approved 225 million ARB over three years, about $215 million at the time, with 76% of votes in favour. Funds sat in the programme’s 3-of-5 multisig. By October 2024, public updates appeared to fall short of the promised fortnightly schedule, and a promised quarterly transparency report had not appeared. A member proposed returning 220 million ARB to the treasury, with the stated aim of accountability rather than ending the programme. A further clawback proposal followed in March 2025.
The lesson is not that the programme was wrong. It is that the DAO had no agreed trigger for pulling funds back, so the only tool was a new political campaign.
Our view: tranche large allocations against named milestones. Require a public report before each tranche is released. State in the original proposal who may propose a recall and on what evidence.
What goes wrong: the question you could not answer
The governance failure that hurts is not a bad vote. It is a question asked two years later that the foundation cannot answer from its own records.
Which passed votes did the council decline to implement, and why? Who held the emergency keys in the first year, and when did they use them? How much of the winning vote came from the labs company, the treasury or early investors? Those questions come from exchanges, acquirers, auditors and regulators. They also matter for securities analysis. The SEC’s March 2026 interpretation describes a decentralized system as one where no person or group holds operational, economic or voting control. If a project promised decentralization, the SEC judges the promise by the project’s own definition.
If the answers live in a deleted chat and a former contributor’s memory, the design on paper does not help. The fix is a governance record owned by the council. Keep each proposal and its outcome, the council’s implementation note, the security council’s actions, and a quarterly snapshot of voting power by holder type.
How Alcyone approaches it
The legal wrapper and constitution belong with counsel. Alcyone works on the commercial layer that governance has to carry. That means token allocation between foundation, labs, investors and community, how much of that supply can vote, and the vesting structures that release it. That work sits within Alcyone’s fundraising and tokenomics advisory. Every plan names an owner for every part, including the governance record after launch.