In a single week of April 2026, roughly $292 million left one protocol through its bridge. No smart contract was broken in the conventional sense, and no private key was guessed. The attacker simply convinced the bridge that a message it had never sent was real, and the bridge released the funds. That incident, the drain of Kelp DAO, was the largest decentralized finance exploit of the year, and it followed a pattern that has held since the first generation of cross-chain bridges launched. This is the first article in a three part series on why bridges keep producing the biggest single losses in crypto, why their dominant design concentrates that risk, and what a structural fix would look like.
What a Bridge Actually Does
A blockchain is a closed system. Bitcoin does not know what happens on Ethereum, and Ethereum cannot read Solana. A cross-chain bridge is the plumbing that lets value appear to move between these separate ledgers. In the most common design, you deposit an asset into a contract on the source chain, that asset is locked, and the bridge mints a representation of it (a wrapped token) on the destination chain. The wrapped token is meant to be redeemable one for one for the original. A separate set of off-chain actors, variously called validators, relayers, oracles, or verifiers, watch the source chain and tell the destination chain when a deposit has happened so the mint can proceed.
For a reader coming from traditional finance, the closest analogy is a depositary receipt. The original asset sits in a vault, and a certificate that trades elsewhere is only as good as the promise that the vault is full and the custodian is honest. The bridge's validators are that custodian. When they can be tricked or compromised, the certificate becomes a claim on nothing.
How the Kelp DAO Drain Worked
On April 18 and 19, 2026, an attacker drained roughly $292 million, about 116,500 units of the staked-ether token rsETH, from Kelp DAO's bridging adapter. The vulnerability was not in the token contract. It was in the messaging layer. Kelp's deployment used the LayerZero protocol to pass cross-chain messages, and the relevant configuration relied on a single verifier, what the industry calls a one of one setup, meaning only one party had to attest that a message was genuine. The attacker, whom LayerZero preliminarily attributed to North Korea's Lazarus Group, compromised the off-chain infrastructure behind that verifier and forged a cross-chain message instructing the bridge to release funds. The bridge did what it was told.
The detail worth holding onto is that the on-chain code behaved exactly as written. The exploit lived in the trust assumption sitting underneath the code: that whoever controlled the verifier was honest and uncompromised. Kelp's team detected the anomaly, paused the relevant contracts, and blacklisted the attacker's addresses, which blocked a follow up attempt to drain a further 40,000 rsETH. Arbitrum's Security Council later froze about 30,766 ether, worth roughly $71 million, that was linked to the theft. Recovery efforts continued for weeks. The point is that the damage was done in minutes by a message, and the cleanup was a slow, human, off-chain process.
Not an Isolated Event
It would be comforting to treat this as one badly configured deployment. The data does not allow that. According to security firm PeckShield, cross-chain bridges suffered roughly $328.6 million in losses across eight major incidents in 2026. Total decentralized finance losses for the year passed $750 million by mid-April, with the Kelp DAO bridge drain and a separate $285 million attack on the Solana exchange Drift as the two largest events. The Drift case was not a bridge exploit (it turned on social engineering of the protocol's signers over several months) but it underlines the same theme: the largest losses now come from the trusted infrastructure around contracts, not from the arithmetic inside them.
The history rhymes too clearly to ignore. The three reference points every bridge engineer knows are the Ronin bridge, drained of roughly $625 million in March 2022 when attackers obtained enough validator keys to approve their own withdrawals; Wormhole, which lost about $320 million in February 2022 after a signature verification flaw let an attacker mint wrapped ether without ever depositing collateral; and Nomad, which bled roughly $190 million in August 2022 after a faulty upgrade made the system treat essentially every message as valid, triggering a copycat free for all. In its 2022 review, Chainalysis estimated that about $2 billion was stolen across 13 separate cross-chain bridge hacks that year, accounting for roughly 69 percent of all crypto stolen. Four years later, the mechanism behind the biggest loss of 2026 is recognizably the same.
The Common Thread
Across Ronin, Wormhole, Nomad, and now Kelp DAO, the failure is not in the token logic. It is in the layer that decides whether a cross-chain message is authentic. Ronin was compromised validator keys. Wormhole was a verification check that could be bypassed. Nomad was an upgrade that broke message validation. Kelp DAO was a single verifier whose off-chain infrastructure was compromised. In each case the bridge faithfully executed an instruction it should never have accepted, because the thing standing between an attacker and the locked collateral was a small, trusted set of off-chain attestations.
This matters because it tells you where to look. The size of a bridge's audit, the reputation of its team, and the formal verification of its contracts are all worth having, but they address the part of the system that has rarely been the problem. The recurring problem is the trust placed in whoever signs off that a message is real. When that trust is concentrated in one party, or a handful, the bridge inherits exactly one point of failure no matter how clean the surrounding code is.
Why This Series, and Where It Goes
For institutions evaluating tokenized assets, stablecoins, or on-chain credit, this is not an abstract security debate. If an asset you hold is a wrapped representation sitting behind a bridge, then your exposure is not only to the issuer and the underlying. It is also to the validator set of every bridge in the path, and to the chance that one forged message turns your holding into an unbacked IOU. The Kelp DAO incident showed that a single message can do in minutes what would take a traditional custodian fraud months to unwind.
The next article examines why the dominant lock and mint design does not just happen to be risky but structurally concentrates failure, turning one breach into simultaneous, multi-chain insolvency. The third looks at what changes when value stays native to a settlement-secure chain instead of being copied behind a bridge. Understanding the anatomy of these exploits is the first step, because the fix is not a better audit of the same architecture. It is a different architecture.
This article is for informational purposes only and does not constitute investment advice.
Mintlayer Web Services provides Bitcoin-native issuance and settlement infrastructure built to keep bridges out of the trust path. Learn more →
