LayerZero Bridge Exploits Expose Design Flaw Affecting $2B in Cross-Chain Losses

Published by James Harris on June 9, 2026June 9, 2026

The $292 million hack of Kelp DAO’s LayerZero-powered bridge in April 2026 did not happen in a vacuum. Cross-chain bridges have now collectively lost over $2 billion to exploits, and the attack vector keeps working for the same structural reason: lock-and-mint bridges concentrate risk in a single smart contract. If the contract holding assets on the source chain is compromised, the wrapped tokens on the destination chain become unredeemable IOUs.

The pattern here is consistent enough to be a design critique, not bad luck. Ronin lost $625 million in 2022. Wormhole lost $320 million the same year. Nomad lost $190 million. Each incident involved a variation of the same failure mode: a custody point that, once breached, could drain everything locked behind it. The Kelp DAO attack fits that lineage precisely, and the fact that it used LayerZero infrastructure matters because LayerZero has been positioned as the messaging layer for a significant portion of DeFi composability. When the infrastructure layer is the attack surface, the blast radius extends beyond a single protocol.

Liquidity-network bridges like Across and deBridge sidestep the wrapped-token problem entirely by settling in native assets using solver or liquidity-pool models, which limits what any single exploit can reach.

The practical hierarchy for users in 2026 runs roughly as follows. Instant-swap aggregators like ChangeNOW carry the smallest contract risk because there is no bridge TVL to drain. Liquidity bridges carry pool-depth risk but avoid the IOU problem. Lock-and-mint protocols like Wormhole offer the broadest chain coverage and developer flexibility, but that breadth comes with a concentrated custody surface that audits can reduce, not eliminate. Stargate and Squid sit in the middle: wide chain coverage, native asset outputs, but dependent on the security assumptions of their underlying messaging layers. Audit count matters less than audit scope; 26 audits on deBridge reflects an iterative security process, but the Kelp DAO attack was also on audited infrastructure.

The selection logic is straightforward: match the bridge type to the actual risk tolerance of the transfer. Large, infrequent transfers across EVM chains favor liquidity networks with native output. Frequent retail swaps across a wide chain range favor aggregators with no custody exposure. Developers building composable cross-chain applications still largely depend on message-passing layers, which means the lock-and-mint risk profile is not going away, it is being priced and managed rather than designed out.