Centralized exchange risks (custody and exposure)

What custody actually means

Custody defines who controls the cryptographic keys, and in crypto, controlling the keys means controlling the assets.

On a centralized exchange, the platform holds those keys. Users do not sign blockchain transactions directly. They interact with the exchange's internal system, where displayed balances are database entries rather than direct on-chain ownership records.

To understand why this distinction matters, see What is a crypto wallet? (custodial vs non-custodial) and Keys, addresses, and seed phrases.

Blockchain transactions still happen, but they are managed in the background, often aggregated across many users. From the blockchain's perspective, individual exchange users are not recognised as owners of those assets. The blockchain recognises only the entity that controls the keys. What users see on screen is the exchange's promise to reflect that control back to them.

Convenience and abstraction

Centralized exchanges handle key management, transaction processing, and account recovery. For users who do not want to manage keys themselves, that removes real friction.

But the same abstraction that simplifies access also removes direct control. The convenience is real, but it comes with a trade-off. That trade-off is structural. It is not a temporary limitation or a feature that will be added later. It is part of how the model works.

Readers who decide they want direct control of their assets can follow How to move crypto off an exchange safely.

Exposure beyond hacking

Exchange risk is often discussed in terms of hacks. Hacking is real, but it is only one way things can go wrong.

Custodial exposure also includes operational failures, insolvency, frozen withdrawals, and regulatory intervention. These are different causes, but they produce the same effect: users cannot act independently at the protocol level. Access depends entirely on the platform's ability and willingness to process requests. That ability can disappear for reasons that have nothing to do with the user.

Irreversibility at a different layer

Blockchains enforce irreversibility at the protocol level. Once a transaction is confirmed, it cannot be undone.

Exchanges add a different kind of layer on top of that. Inside the exchange, some actions can appear reversible. Balances can be adjusted, accounts restricted, and withdrawals delayed or stopped. That flexibility belongs to the exchange, not to the user.

When an exchange fails, there is no blockchain-level recovery path for users who never controlled the keys. The blockchain will show the exchange as the entity that controlled those keys, because it was. The internal promise that those funds belonged to users existed entirely outside the protocol, and the protocol cannot enforce it.

Trust as a single point of failure

Using a centralized exchange means placing trust in something the protocol cannot see.

That trust is not in cryptography. The cryptographic layer still works as intended. It is trust in the exchange's operations, governance, and solvency, in the company remaining functional, solvent, and willing to honour withdrawals when requested. Whether those conditions hold is not something the protocol tracks or guarantees.

If that trust turns out to be misplaced, the protocol offers no remedy. It already did its job by processing the signatures the exchange submitted. What happens between the exchange and its users is outside its scope entirely.