LayerZero

Ryan Zarick publicly introduced LayerZero in September 2021 with Bryan Pellegrino and Caleb Banister as the other founders of LayerZero Labs. The design did not ask destination chains to understand every source chain. An Ultra Light Node received a block header from an Oracle and a transaction proof from a Relayer, then checked whether the proof belonged under that header.

This split reduced the work done on-chain, but it did not remove trust. The selected Oracle and Relayer had to remain independent; if they colluded, the destination could accept a false message. LayerZero’s own explanation said the worst case fell back to the security of the Oracle. Applications could choose the pair, so security was already partly an application decision.

Stargate launched in March 2022 as the first large product built on that transport. Its unified pools and Delta algorithm let users move native assets without wrapping them or waiting for a second transaction. That success made a useful distinction visible: LayerZero carried instructions, while Stargate managed liquidity and user balances.

The same year also exposed how much had to remain replaceable. An anonymous disclosure identified a pathway-griefing issue in ULNv1. LayerZero deprecated that library, patched the problem in ULNv2 and kept the endpoints in place. Immutability applied to deployed components, while applications could opt into a newer Message Library.

LayerZero announced V2 in December 2023 and put it live on more than twenty chains on January 29, 2024. The Oracle role became a broader class of Decentralized Verifier Networks. The Relayer’s two tasks were split: DVNs attest to data, while permissionless Executors deliver verified messages and supply destination gas.

An OApp can require every DVN in one list and a threshold from another. It can mix LayerZero Labs, cloud operators, native bridges or other verification systems. The Endpoint and each released Message Library remain immutable, but an application owner or delegate can change which library, peers and DVNs a channel trusts.

That transfer of choice was the point of V2, not an accidental admin feature. It let a high-value token demand several independent attestations and a lower-value application buy less verification. It also meant the LayerZero name alone no longer described the security of a particular route.

Before the June 20, 2024 claim, LayerZero reviewed nearly six million interacting addresses for Sybil behavior. The final claim asked recipients to pay $0.10 per ZRO to Protocol Guild, an unusual condition that produced congestion and arguments over whether a required payment should be called a donation. The condition belonged to that distribution event; it was not a recurring message fee.

The initial allocation described one billion ZRO divided among community, strategic partners, contributors and a repurchased 40-million-token bucket. The project’s 2026 account later said the FTX legal settlement returned that bucket to strategic partners and changed their allocation to 312 million. This was another separation: protocol history, corporate settlements and token allocation could affect one another without changing how an Endpoint verifies a packet.

The breach behind the April 18, 2026 KelpDAO loss began weeks earlier. According to LayerZero Labs’ final report, an attacker socially engineered a Labs developer, stole session access and altered internal RPC nodes so monitoring received honest answers while the Labs DVN received false ones. A denial-of-service attack then removed the external RPC provider from the signing path.

KelpDAO’s rsETH channel had previously moved from two required DVNs to a 1-of-1 configuration using only the LayerZero Labs DVN. Once that DVN signed the forged message, no independent verifier remained to disagree. The destination released 116,500 rsETH, worth about $292 million at the time.

Calling this only a bad application setting would omit the compromised Labs infrastructure. Calling it an Endpoint hack would omit that the on-chain protocol followed the configuration it had been given. The loss joined an operational compromise to an application choice—the exact boundary V2 had placed in the hands of each OApp.

LayerZero Labs rebuilt the affected environment and said its DVN would no longer serve as the sole required attestor on a channel. The protocol did not revoke an OApp’s ability to choose its own stack. Instead, Labs changed which configurations its own DVN would support, while default V2 routes moved to at least 3-of-3 verification.

By August 4, active V1 default routes had moved from ULNv2 to ULNv301 with the same minimum. Applications pinned to the older library remained outside that default migration and faced a December 15, 2026 deadline to migrate or run their own Relayer. LayerZero’s lasting result is therefore specific: applications retain the choice V2 promised, but the main operator stopped treating every possible choice as one it should help execute.