What is a DAO, really, and what they keep getting wrong
DAOs were sold as on-chain corporations. After ten years of evidence — from The DAO to ConstitutionDAO to ENS — the working definition is narrower, and more interesting.
On 17 June 2016, an attacker drained 3.6 million ETH from a smart contract called The DAO, then worth roughly $60 million, by recursively calling its splitDAO function before the balance updated. The attack forced Ethereum’s first contentious hard fork, split the chain into ETH and ETC, and froze a generation of legal questions about what a "decentralised autonomous organisation" even is. Almost ten years later, the term DAO covers everything from a Discord with a multisig (Friends With Benefits) to a $200 million protocol treasury voted on by 80,000 token holders (Uniswap), to a fork-of-a-fork buying a copy of the US Constitution in 47 hours (ConstitutionDAO). Most of these things are not what the 2014 whitepaper described, and pretending otherwise has cost a lot of money.
What is at stake when you misread a DAO is straightforward: people send treasury funds to multisigs they cannot reclaim, vote on proposals whose execution they don’t understand, and pay legal fees for entities that have no clear jurisdictional status. The Coin Center case work, the CFTC’s 2022 Ooki DAO enforcement, and Wyoming’s DAO LLC statute are three data points that should be required reading before anyone joins their first governance call. This piece tries to define the term narrowly enough to be useful — and to identify the failure modes that have shown up in the wild.
The original definition, and why it is mostly wrong
Before any of this matters, you have to read the source. The phrase "decentralized autonomous organization" was popularised by Vitalik Buterin’s May 2014 blog post, which proposed a taxonomy: a DAO is "an entity that lives on the internet and exists autonomously, but also heavily relies on hiring individuals to perform certain tasks that the automaton itself cannot do." The key word is autonomous: the entity is supposed to run by its own code, not by the discretion of operators.
Almost no organisation that calls itself a DAO today meets this definition. Uniswap’s governance can pause fees but cannot upgrade the protocol without a code deployment by humans. ENS’s DAO votes on treasury allocation but the technical work is done by ENS Labs, a Cayman Islands foundation. MakerDAO — now rebranded to Sky — runs its peg-stability mechanism algorithmically, but core changes still require human-authored executive votes. The autonomous part faded; the "collectively governed treasury and protocol parameters" part stuck.
A working taxonomy — five things people call DAOs
| Type | Example | Treasury size | What it actually does |
|---|---|---|---|
| Protocol DAO | Uniswap | $3.1B UNI + $19M USDC | Votes on fee switch, grants, deployment to new chains |
| Service DAO | MakerDAO / Sky | $1.2B in MKR/SKY + reserves | Manages stablecoin collateral, sets risk parameters |
| Investment DAO | MetaCartel Ventures | ~$15M | Pools capital, votes on investments, distributes returns |
| Social DAO | Friends With Benefits | $2M treasury | Gates a community, funds events, issues a token |
| Single-purpose DAO | ConstitutionDAO | $47M raised in 47 hours | Pooled capital to bid at Sotheby’s, refunded after losing |
Each row in this table has a different legal exposure, governance architecture, and failure mode. A Protocol DAO is mostly a Cayman or BVI foundation with a token-weighted vote attached. A Social DAO is mostly a Discord with a multisig. Calling both "DAOs" is the term’s most expensive ambiguity. The DeepDAO dashboard tracks roughly 13,000 of these entities; fewer than 200 have ever passed a proposal in a calendar year.
Governance theatre — the participation problem
Token-weighted voting sounds democratic until you look at the turnout. Uniswap’s average proposal sees voting power from fewer than 20 addresses controlling more than half of the eligible UNI. ENS’s most-contested proposal of 2024 had 28 delegates casting 73% of the weight. The Snapshot data is public: most DAOs are oligarchies with a long tail of small holders who don’t vote because the gas cost (on-chain) or the time cost (off-chain) exceeds the influence they have.
The pattern that has emerged is delegation: small holders assign their voting power to named delegates who promise to vote a particular way. ENS, Uniswap, Compound, and Optimism all use this model. It works in the sense that it gets votes counted; it fails in the sense that the median holder has no real say. The 2024 DAO Times participation report put median voter participation across the top 50 DAOs at 3.7% of circulating tokens.
Legal entities — Wyoming, Cayman, Marshall Islands
For a long time, DAOs operated in legal limbo. That changed in stages. Wyoming passed its DAO LLC statute in March 2021, allowing a DAO to wrap itself in a US limited liability company structure. The Marshall Islands’ Non-Profit Entities Act, amended in 2022, created the "DAO LLC" offshore version, used by Shipyard and others. Cayman foundations remain the structure of choice for protocols that need an entity to sign contracts and hold IP.
The CFTC’s September 2022 action against Ooki DAO for operating an unregistered futures platform — and serving the lawsuit by posting in the project’s Discord — was the moment US enforcement clarified that token-voting alone does not insulate participants from liability. The final order in June 2023 held token holders jointly liable as "an unincorporated association." If you are a US-based DAO member, that case sets the floor.
What DAOs are actually good at
Strip away the autonomy claim, the legal uncertainty, and the participation theatre, and there is a real thing left. DAOs are unusually good at three specific tasks: pooling capital quickly without a bank, distributing a treasury to grant recipients in public, and changing protocol parameters where the change set is small and well-defined. ConstitutionDAO raised $47M in 47 hours through a single contract on JuiceboxDAO; no traditional vehicle moves that fast. Optimism’s RetroPGF rounds have distributed over $200M to public goods builders through a structured voting process that would be impossible to administer in a normal grants program.
- Capital formation: a smart contract collects ETH or USDC from anyone, with automatic refunds if a quorum or deadline isn’t met.
- Treasury transparency: every disbursement is on-chain, every signer is named, the whole ledger is readable on Etherscan.
- Parameter governance: when a vote outcome maps directly to a function call (set fee tier, add collateral type), DAO governance is faster and cheaper than corporate process.
What they keep getting wrong
The three repeated failure modes are worth naming. First, multisig drift: a DAO is created with the promise of full on-chain voting, then quietly delegates execution to a 5-of-9 multisig that operates by Telegram. The MakerDAO Endgame restructuring, the Wonderland-DAO treasury management episode of 2022, and the Frax Finance core-team multisig are all examples. The DAO becomes a marketing label for what is functionally a small executive committee.
Second, vote buying through bribes. Convex and Aura built entire businesses around bribing veCRV and veBAL voters to direct emissions toward specific pools. The bribes are technically legitimate — they are denominated transactions to known addresses — but the result is that protocol parameters are set by the highest bidder rather than the protocol’s stated mission. Votium processes tens of millions of dollars in bribes per epoch.
Third, treasury mismanagement. The median DAO treasury is denominated 70%+ in its own governance token. When the token drops, the treasury drops with it, and the DAO discovers it cannot pay its contributors. ENS, Compound, and Uniswap diversified into stablecoins after 2022; most smaller DAOs did not, and several have effectively run out of operational runway in their own collapsing native asset. Our tools page includes a treasury composition tracker for the top 100 DAOs by AUM.
The honest definition
A DAO in 2026 is, almost always, a token-gated coordination system that uses on-chain or hybrid voting to allocate a treasury and adjust the parameters of an associated smart-contract protocol, usually backed by a foundation in a friendly jurisdiction. It is not autonomous in the 2014 sense. It is not a corporation in the Delaware sense. It is a new organisational form with real strengths in transparency and capital formation, and well-documented weaknesses in participation, security, and legal clarity. Treat it as that, read the proposals on Tally before you vote, and check the multisig signers before you trust the "decentralised" label. The interesting DAOs of the next five years will be the ones that admit what they are.
A short history — the four eras of DAO experimentation
Reading the history in periods makes the present moment legible. The first era, 2014-2016, was theoretical and ended with The DAO hack. The second, 2017-2019, was dominated by Moloch DAO and MetaCartel — small, multisig-led grant-making groups that demonstrated DAOs could function without on-chain enforcement of every decision. The third era, 2020-2022, was the "DAO summer" — every protocol launched a governance token, every grant program rebranded as a DAO, and the assumption was that token-weighted voting would scale to billions in treasury. It did not. The fourth era, post-2022, is consolidation: large protocol DAOs (Uniswap, ENS, Aave, Compound) running on professional delegate networks, with most other "DAOs" quietly reorganising as foundations or simply going dormant.
This periodisation matters because each era’s failure mode informed the next era’s design. The DAO hack produced the modern norm of timelocks and circuit breakers on any contract that holds significant capital. The Moloch era’s success with simple ragequit-and-multisig governance was the template for investment DAOs. The DAO summer’s failures — collapsing tokens, no-show voters, captured governance — produced the current delegate-based hybrid model. Our events calendar tracks the major governance votes worth watching, and the market page charts governance-token treasury exposure for the top protocols.
Tooling — the working software stack in 2026
If you are actually going to participate in a DAO, knowing the tooling is as important as understanding the theory. Snapshot is the off-chain signalling layer — proposals are posted, token holders sign messages with their wallets, results are aggregated and either taken as advisory or piped into an on-chain executor. Tally is the front-end of choice for on-chain voting through the Governor framework that Compound popularised. Safe (formerly Gnosis Safe) is the multisig wallet that almost every DAO treasury actually sits in. Zodiac by Gnosis Guild builds modular extensions on top of Safe — Reality.eth oracles, delay modules, scope guards. Aragon remains the all-in-one option for new DAOs that want a single stack rather than assembling components.
For grant-making, Karma and Charmverse handle the application-and-review workflow most DAOs converged on after the ad-hoc forum-thread approach scaled badly. For delegate research, Boardroom and Agora aggregate voting history across protocols. The interesting common thread is that none of these were built by the DAOs they serve; tooling has consolidated into a small number of specialised providers, which is itself a form of centralisation worth noticing. Our tools page tracks the dashboards we use to monitor governance activity in real time.