The reason a multisig overrides the vote it claims to execute

A DAO voted, but execution ran through a multisig of a few trusted signers who had to enact the result on-chain. That step was described as mere implementation. In truth it was a control point: the signers could delay indefinitely, execute selectively, or decline entirely, and the community had no way to force their hands. A vote that only becomes real when a few people choose to enact it is not binding; the multisig holds the final say behind the language of carrying out the community’s will.

What everyone sees

A voter sees a passed proposal handed to a multisig for execution and reads a clerical step: the decision is made, the signers just implement it. The framing implies obligation. The voter treats execution as automatic, not discretionary, and does not see that signers who can delay, alter or refuse hold a veto in practice, making the binding vote binding only if they choose to bind themselves to it.

What is actually happening

Research on principal-agent problems shows that whoever controls execution controls outcomes when they cannot be compelled. A multisig that must act for a vote to take effect is an agent with discretion, not a passive instrument, and absent an enforcement mechanism the signers can substitute their judgement for the vote. The gap between deciding and executing is where power actually sits, and the multisig occupies it.

Why it stays hidden

The hidden mechanism is discretionary execution masquerading as implementation. The vote produces a result; the multisig decides whether that result happens. Framing the step as clerical hides the veto it contains. The voter believes the decision is final, while the signers hold the power to delay, reshape or ignore it, and the community’s will is real only at their discretion.