A DAO lets token holders vote on how a treasury gets spent instead of a CEO deciding alone, but the record shows that voting power concentrates fast and most holders never show up to vote at all.
What is a DAO? Strip away the jargon and it's a group that pools money and makes decisions on-chain instead of through a board of directors. No CEO signs off on spending. No legal entity issues a memo. Instead, a smart contract holds the treasury, and anyone holding the group's governance token can propose how that money gets used and vote on whether it happens. That's decentralized autonomous organization governance in one sentence, and it's also where things get complicated fast.
The idea took off after The DAO launched on Ethereum in 2016, raised roughly $150 million in ether from thousands of investors, and then got drained of about $60 million through a code exploit within its first month. Ethereum's response, a hard fork that split the chain into Ethereum and Ethereum Classic, is still the most consequential governance failure in crypto history. It's also the best answer to anyone who thinks DAOs are a solved problem. They aren't. They never really have been.
Most DAOs today run on a simple loop. Someone with enough tokens or reputation submits a proposal, usually through a platform like Snapshot or Tally. The community debates it on a forum, often Discourse or Discord. Then token holders vote, and if the proposal clears a quorum threshold, a multisig wallet or an on-chain execution contract carries it out. Uniswap's DAO governs a treasury that has held billions of dollars in UNI tokens through exactly this process. MakerDAO, which oversees the DAI stablecoin, runs continuous executive votes where MKR holders approve changes to collateral types and risk parameters, sometimes weekly.
That sounds tidy. It rarely stays tidy. Snapshot votes are typically off-chain and gasless, meaning they're signals of intent rather than binding transactions, so someone still has to execute what the community decided. That gap between voting and execution is where trust has to fill in for code, which undercuts the entire pitch of removing intermediaries.
Governance Tokens Decide Who Actually Rules
Here's the part most explainers gloss over: a DAO's governance token is not a democracy ticket. It's a share, and shares concentrate. Uniswap's UNI token was distributed to early users in 2020, but a large share of voting power still sits with venture funds and the founding team through allocations and vesting schedules. A DAO voting token gives you influence proportional to how much you hold, not proportional to how much you care, and that single design choice explains most of what goes wrong downstream.
Compound, the lending protocol, learned this the hard way in 2021 when a bug in a contract upgrade mistakenly distributed roughly $90 million worth of COMP tokens to users. Fixing it required governance votes, and getting a proposal through fast enough to matter meant courting the same large token holders whose outsized stakes are the thing critics point to as the model's core weakness.
Voter apathy compounds the concentration problem instead of offsetting it. A widely cited study of major DAOs found that a small handful of addresses cast the majority of votes across proposals, while most token holders never vote on anything. Low turnout means the whales who do show up decide outcomes almost by default. It isn't a fringe issue. It's closer to the normal operating condition of on-chain governance.
Treasury Attacks Are the Failure Mode That Actually Costs Money
Bad voter turnout is a slow leak. Treasury attacks are a rupture. Beanstalk Farms, a stablecoin protocol, lost roughly $182 million in April 2022 when an attacker took out a flash loan, used it to instantly acquire enough governance tokens to pass a malicious proposal, and drained the treasury in a single transaction before anyone could react. The proposal and the theft happened in the same block. There was no debate period long enough to stop it because the attacker never needed one.
Beanstalk is the clearest illustration of a structural flaw in flash-loan-vulnerable governance: if voting power can be borrowed for the length of one transaction, quorum and majority rules protect against almost nothing. Many DAOs responded by adding timelocks, delays between a vote passing and execution, specifically so a community has a window to notice and counter an attack. That fix is a patch on a design flaw, not a solution to it, and newer DAOs still get caught without one.
Frankly, this is the tension nobody selling DAOs to founders wants to say out loud: the more decentralized the voting, the slower and more exploitable the system gets, and the more concentrated it gets to stay fast and secure, the less it resembles the flat, leaderless structure DAOs are marketed as.
What This Actually Means If You're Building One
If you're a founder weighing whether to launch a DAO instead of a normal company, the honest pitch is narrower than the hype suggests. DAOs work well for coordinating a treasury among people who already broadly agree on goals, like funding public goods or managing a protocol's parameters, where ArbitrumDAO and its roughly $2 billion in prior treasury allocations is a working example of large-scale on-chain fund governance. They work badly as a substitute for fast, accountable executive decision-making, which is why even crypto-native teams increasingly keep day-to-day operations off-chain and reserve DAO votes for major treasury or protocol changes.
The vote-then-execute loop, the concentration of governance tokens among early holders and funds, and the ever-present risk that someone borrows enough voting power for one transaction to move a treasury, these aren't edge cases to design around later. They're the actual mechanics of how decentralized autonomous organizations function right now, in 2026, after a decade of iteration. Snapshot, Tally, and timelocked multisigs have made the process more legible than it was in 2016. They haven't made it safe by default.
Anyone asking what a DAO is deserves a straight answer: it's a treasury governed by a token-weighted vote, and every real-world DAO failure since The DAO itself traces back to that one sentence. Beanstalk didn't get hacked because the code had a typo. It got hacked because the governance model let anyone rent a majority for a few seconds. That's not a bug you patch once. It's the trade-off you're signing up for the moment you decide a smart contract should hold the money instead of a person.
Also read: What Is a Perpetual Futures Contract and Why Funding Rates Decide Who Gets Liquidated • What Is a Rug Pull in Crypto and How to Spot One Before You Buy • What Is Agentic AI and How Do Autonomous AI Agents Actually Work
