Crypto Tornado Cash: The DAO Governance Question

Prepared by the editorial team. Updated August 31, 2026.

Crypto Tornado Cash carried a governance token, TORN, and an associated decentralised autonomous organisation, and that fact complicates the tidy story usually told about the protocol. A governance system exists so that participants can change something, which is a claim about human decision making rather than about fixed code. Understanding which decisions the votes could actually reach is what turns a slogan into an accurate description.

What did the DAO actually control?

Governance controlled the parts of the system that were designed to be changeable: the treasury holding TORN, the parameters of peripheral contracts, the allocation of funds to work the community wanted done, and the governance rules themselves. It did not control the pool contracts that hold and release deposits, because those were deployed without any function a vote could call.

A token-holder governance system is a contract like any other. Holders lock or delegate tokens to gain voting weight, someone submits a proposal in the form of executable instructions, a voting period runs, and if the proposal passes and clears any delay, the governance contract executes those instructions. The system can only ever call functions that some other contract exposes to it.

That last point is the whole answer to the question. A governance contract has reach exactly as far as the permissions other contracts grant it, and no further. Where a target contract has no owner, no admin role and no upgrade hook, there is no door for governance to open. The pool contracts fell in that category by construction.

What remained was still meaningful. Treasury decisions direct real money, and changes to the governance rules themselves determine who gets to decide next time. None of that is trivial, and all of it involves identifiable people voting.

Why does a governance layer strain a claim of full autonomy?

Because autonomy claims that no party can alter the system, while a governance layer exists to let a party alter part of it. Both statements can be true when each is scoped to its own components, but the unscoped version is contradictory: a system cannot simultaneously be beyond anyone’s control and be steered by a vote.

The tension is not a flaw in the design so much as a flaw in the way the design gets described. Engineers generally understood the architecture as layered, with immutable pools at the base and mutable coordination around them. Public summaries flattened those layers into a single adjective, and the adjective chosen was usually the strongest one available.

Flattening has consequences beyond imprecision. If a reader believes nothing about the system can be changed by anyone, then the existence of a treasury, a proposal process and a set of large holders undermines trust in every other claim made about the project.

Why does this layer attract legal attention?

Because law looks for people who make decisions, and a governance system is a record of people making decisions. Immutable code offers nothing to attach a duty to, while a proposal, a vote and an executed treasury transfer are documented acts by identifiable participants. That is why analysis of these cases moves quickly from the pools to the governance layer.

Several distinct legal questions arise from the same structure. One is whether the arrangement amounts to an entity, such as a general partnership, that could bear obligations. Another is whether particular participants took actions satisfying the elements of a specific offence. A third is how the tokens themselves are characterised. These are independent questions with independent answers.

The public record shows the pattern. When Treasury announced its 2022 designation of the mixing service, the action reached names and addresses. The later United States prosecution of Roman Storm concerned conduct by a person, and in August 2025 a jury convicted him on one count of conspiracy to operate an unlicensed money transmitting business while deadlocking on two others. A deadlock is not an acquittal, and those counts are set for retrial in April 2027.

None of this establishes a general rule about governance participation, and no court has issued a decision that resolves the question for all arrangements of this kind. The law here is genuinely unsettled, and anyone whose own participation raises the question should consult qualified counsel rather than infer a rule from reported outcomes.

How can you reconstruct a governance record from on-chain data?

You reconstruct it by locating the governance contract, reading its proposal and vote events in chronological order, separating proposals that executed from those that did not, measuring how concentrated the deciding votes were, and checking which contracts each executed proposal could actually call. The result is a factual record rather than a summary of what people said they intended.

Step 1: Identify the governance contract itself

Find the address of the contract that records proposals and votes, and confirm it against more than one independent reference before you rely on it. Governance systems frequently span several contracts, so establishing which one holds the authoritative record comes first.

Step 2: Read the proposal events in order

Pull the event logs the governance contract emitted when proposals were created, voted on and executed, and arrange them chronologically so the record reads as a sequence rather than a list. Ordering reveals reactions and reversals that a table of outcomes hides.

Step 3: Separate proposals that executed from those that did not

Distinguish proposals that reached execution from those that failed, expired or were withdrawn, because a discussion thread often records intentions that the chain shows were never carried out. Commentary tends to cite the ambition rather than the outcome.

Step 4: Measure how concentrated the voting was

Compare the voting weight that decided each outcome against the total supply eligible to vote, since a proposal passed by a handful of large holders tells a different story from one with broad participation. Concentration is a property of the record, not an accusation.

Step 5: Check what each executed proposal could actually touch

Trace which contracts an executed proposal was able to call, because that boundary defines the real reach of the governance system regardless of how the proposal was described. This is the step that separates governance in name from governance in effect.

Does holding a governance token create responsibility for a protocol?

Holding a token is not by itself an act of governance, and most legal analysis focuses on conduct rather than on balances. Voting, proposing, executing and taking a defined role are conduct. Whether any of that conduct creates liability depends on the jurisdiction, the facts and the statute involved, and the question remains unsettled.

The reason this is unresolved is structural. Company law developed alongside registries, formation documents and defined offices, so it can identify who owes what to whom. A governance arrangement with pseudonymous participants and no formation document does not map onto those categories.

Two practical points survive the uncertainty. First, the more a participant looks like someone exercising ongoing control or receiving compensation for a role, the more the analysis resembles familiar questions about operators and managers. Second, a passive holder who never votes is in a different factual position from an active proposer, and blending the two produces bad reasoning in both directions.

Governance reach layer by layer

The table maps each layer of the system against whether a token vote could change it. It is a general illustration of how governance permissions work in layered protocol designs rather than a legal assessment of any particular arrangement.

Layer	Could a token vote change it?
Core pool contracts	No, because they expose no owner or admin function for governance to call
Governance treasury	Yes, since allocation of held tokens was the central use of the vote
Peripheral contract parameters	Yes, where a contract explicitly granted governance permission to set them
Voting rules and thresholds	Yes, and changes here alter who can decide everything else afterwards
Hosted interfaces and domains	No, because these sat with off-chain parties beyond any contract’s reach
Individual relayer participation	No, since each relayer independently decided whether to run the service

The rows divide into three groups: things governance controlled directly, things it could never touch because no permission existed, and things that were never on chain at all. Most confused claims about DAO power come from moving an item between those groups without noticing.

Frequently asked questions

Can a DAO be wound down the way a company can?

Not cleanly. A vote can disable governance functions or empty a treasury, but it cannot delete deployed contracts that have no removal mechanism, and it cannot dissolve whatever legal status the arrangement may have acquired in a given jurisdiction. Winding down the visible activity and ending the legal existence are separate problems.

What happens when almost nobody votes?

Low participation concentrates effective control in whoever does vote, so a small group of large holders can carry proposals that most holders never see. Quorum rules are meant to prevent this, but they can also freeze a system entirely when turnout falls below the threshold, leaving decisions impossible either way.

Is a passed proposal automatically binding on anyone?

Only to the extent the contracts execute it. A proposal that calls a function the governance contract controls takes effect automatically, while a proposal expressing a policy, a request or a statement of intent depends entirely on people choosing to act on it. The two kinds are often listed together and read as equivalent.

Does holding a treasury make a DAO look more like a company?

It is one of the factors that invites the comparison, alongside pooled assets, decisions taken by vote, and people acting in recognised roles. No single factor decides how a court or regulator will characterise an arrangement, and the analysis is jurisdiction specific, which is why anyone actually involved should seek qualified counsel.