Cross-chain bridges lost $340.7 million across 14 exploits in 2026 through September, according to aggregated data from Chainalysis and PaNews. The $292 million Kelp DAO breach in April accounted for 86% of that total — attributed to North Korea's Lazarus Group, it exploited a single-verifier con...
"Historically, legacy bridges have lost billions due to insecure infrastructure, while in-house builds are slow and expensive and institutions' proprietary networks can't earn the trust of their peers." — Johann Eid, Chief Business Officer, Chainlink Labs
Cross-chain bridges lost $340.7 million across 14 exploits in 2026 through September, according to aggregated data from Chainalysis and PaNews. The $292 million Kelp DAO breach in April accounted for 86% of that total — attributed to North Korea's Lazarus Group, it exploited a single-verifier configuration on a LayerZero-based bridge, draining 116,500 rsETH across 20 chains and forcing market freezes at Aave, SparkLend, and Fluid.
The four dominant interoperability protocols — Chainlink CCIP, LayerZero, Wormhole, and Hyperlane — have responded to this loss data by diverging further in architecture and target market. LayerZero processes 75% of all cross-chain message volume across 165 chains but faces a lawsuit from Kelp DAO over the single-verifier configuration it allegedly endorsed. Wormhole's Portal Bridge holds nearly $3 billion in TVL with a rigid 19-of-19 Guardian multisig. Chainlink, which launched CCIP 2.0 on September 28 at Sibos 2026, now secures $84 billion in cross-chain token value and has embedded compliance controls targeting regulated institutions. Hyperlane connects 140+ chains with modular security modules but holds $91.6 million in TVL.
This report compares the security architectures, trust models, market positions, and loss records of these four protocols as the interoperability market segments by user type rather than consolidating around a single standard.
Cross-chain bridges remain the most capital-intensive attack surface in decentralized finance. Through September 2026, 14 bridge-related exploits resulted in $340.7 million in confirmed losses. This figure exceeds the approximately $300 million in total bridge losses recorded across all of 2024.
The distribution is heavily concentrated. The Kelp DAO exploit on April 18 accounted for $292 million — roughly 86% of the year's total. Excluding that single event, the remaining 13 exploits averaged $3.7 million each. In July, AFX Trade and Verus lost a combined $31.5 million in back-to-back bridge attacks within 24 hours. Hyperlane suffered a $2.5 million exploit in April.
The data points to a pattern: the highest-value losses in 2026 stemmed from social engineering attacks on verification operators, not from smart contract logic failures. According to a DEV Community post-mortem analysis of the Kelp DAO incident, the bridge code itself had been audited. The failure was architectural — a configuration choice, not a code bug.
This distinction matters for protocol comparison. The question is no longer whether bridge code can be written securely, but whether bridge architectures can prevent deployers from selecting insecure configurations.
On April 18, 2026, attackers drained approximately 116,500 rsETH — worth $292 million — from Kelp DAO's cross-chain bridge. The attack represented 18% of rsETH's circulating supply and disrupted markets across 20+ chains.
Attack sequence, per Chainalysis post-mortem:
Aave, SparkLend, and Fluid froze rsETH-related markets immediately. Kelp DAO's parent entity, Evercrest Technologies, subsequently filed suit in the Supreme Court of British Columbia against LayerZero Labs and co-founder Bryan Pellegrino. The filing alleges LayerZero staff reviewed and endorsed the single-verifier configuration.
Pellegrino responded publicly: "[N]obody should be relying on sole DVN." He later called the suit "meritless."
The dispute encapsulates the central tension in configurable bridge security: when a protocol offers modular verification and a deployer selects a minimal configuration, liability is ambiguous. The court proceeding, still active as of September 2026, may set precedent for how shared-responsibility security models are treated under contract law.
The cross-chain interoperability market has consolidated around four dominant protocols. Each embodies a distinct philosophy on the tradeoff between configurability and default security guarantees.
CCIP 2.0, launched September 28 at Sibos 2026, does not compete on chain count. It targets regulated asset transfers where compliance and multi-layered verification are prerequisites. A tokenized bond issuer, for example, can require its own compliance team to co-sign every cross-chain movement alongside Chainlink's DON. The faster-than-finality transfer option allows destination-chain delivery before full source-chain finality, with full finality remaining the default.
LayerZero's V2 protocol treats security as a composable, application-chosen service. This design enabled rapid chain coverage — 165+ networks, more than any competitor — and dominance in DeFi messaging volume. The tradeoff became concrete on April 18: a single deployer's configuration choice resulted in the largest bridge loss of 2026. Whether this represents a protocol design failure or a deployment failure is now a legal question.
Wormhole's approach is the most rigid: 19-of-19 means every Guardian must validate every message. This eliminates single-point-of-failure risk by construction but limits throughput and makes adding new chains dependent on Guardian coordination. The NTT framework adds application-level safety features — rate limits, pause controls — that operate independently of the messaging layer.
Hyperlane occupies the permissionless end of the market. Any chain can deploy Hyperlane without permission, and applications select their security model from a menu of ISM primitives. The ZK-proof option offers the strongest cryptographic guarantee available in any current bridge protocol, but adoption of ZK ISMs remains limited due to cost and complexity. Most deployments use multisig ISMs.
| Metric | CCIP | LayerZero | Wormhole | Hyperlane | |--------|------|-----------|----------|-----------| | Value secured/TVL | $84B | $370M (Stargate) | ~$3B (Portal) | $91.6M | | Chains | 30+ | 165+ | 30+ | 140+ | | Message volume share | <5% | 75% | ~20% | <5% | | Default security floor | High (16-node DON) | Low (deployer-chosen) | High (19/19 multisig) | Low (deployer-chosen) | | Compliance tooling | Native (ACE) | None | None | None | | Primary users | Institutions | DeFi protocols | Token issuers | Modular chains | | 2026 losses | $0 | $292M | $0 | $2.5M |
Circle's Cross-Chain Transfer Protocol handles USDC-specific transfers through a burn-and-mint model with Circle as the sole verifier. This is the tightest trust assumption in the market — backed by the issuer's own solvency rather than a decentralized validator set. Circle Gateway holds $63.75 million in TVL across 7 chains. CCTP V2 is not a general-purpose messaging protocol, but it demonstrates a viable single-issuer security model for the specific case where the bridge operator and the asset issuer are the same entity.
The four protocols have segmented into distinct market tiers rather than competing head-to-head.
LayerZero dominates raw throughput: 75% of cross-chain message volume, $293 million in average daily transfers. Its OFT framework, with 733+ tokens deployed, has become the de facto standard for DeFi token portability. Stargate Finance, its flagship bridge, holds $370 million in TVL across 26 chains.
Wormhole leads in bridge-held capital: Portal Bridge's approximately $3 billion TVL exceeds all other messaging-protocol bridges combined. Its 1 billion cumulative transactions reflect consistent retail and institutional usage, though its 20% message share suggests lower-frequency, higher-value transfers.
Chainlink CCIP leads by a metric of its own construction: "value secured" — the total token value relying on its infrastructure, currently $84 billion. This figure dwarfs its actual transfer volume. The metric counts all assets whose cross-chain operations depend on CCIP, regardless of whether those assets move on a given day. Fee data at the CCIP level remains opaque; Chainlink does not publish granular revenue figures for its cross-chain product.
Hyperlane serves the long tail: 140+ chains, many of them smaller or application-specific. Its permissionless deployment model attracts modular blockchain ecosystems that cannot or will not negotiate integrations with larger protocols.
The institutional lineup around each protocol provides a proxy for how regulated capital evaluates cross-chain risk.
CCIP integrations: ANZ Bank, Fidelity International, Deutsche Börse's Crypto Finance unit, SBI Digital Markets, BitGo, Coinbase (oracle infrastructure for tokenized stocks), State of Wyoming (FRNT stablecoin). Infosys — a $75 billion market-cap IT services company — announced a partnership on September 22 to develop CCV and compliance infrastructure.
LayerZero: No major regulated financial institution has publicly committed to LayerZero for primary asset transfer infrastructure. Its dominance is in DeFi-native protocols — Stargate, cross-chain lending, OFT deployments.
Wormhole: Adopted by several token issuers for NTT deployments. Its Guardian set includes institutional operators, but public partnerships with traditional financial institutions are limited.
Hyperlane: Primarily adopted by modular blockchain projects and application chains. No institutional finance partnerships announced.
The pattern is consistent: institutions select protocols with higher default security guarantees and compliance tooling, even at the cost of chain coverage. LINK token price reflected this dynamic — trading at $14.40 on September 28 before the CCIP 2.0 announcement, rising to $15.43 by session close (up 10%), and approaching $16 by September 30 for an 18% weekly gain.
The cross-chain bridge market after the Kelp DAO breach and CCIP 2.0 launch is a three-tier structure. LayerZero owns retail and DeFi throughput at 75% message share across 165 chains. Wormhole holds the most capital in bridge contracts at roughly $3 billion TVL. Chainlink claims the institutional tier at $84 billion in secured value with compliance tooling no competitor currently matches. Hyperlane occupies the permissionless frontier.
Before April 2026, the debate around bridge security was largely theoretical. After $292 million disappeared through a documented-but-permitted single-verifier configuration — and a lawsuit followed — the conversation shifted to default security floors, deployer accountability, and the legal allocation of responsibility in modular security architectures.
Each protocol's response to the Kelp DAO breach reveals its strategic priorities. Chainlink added institutional verification layers. LayerZero faces litigation over its configurability model. Wormhole's rigid 19-of-19 Guardian model, while expensive to operate, has avoided any exploit on its current architecture. Hyperlane's modular ISM approach offers the strongest theoretical security (via ZK proofs) but the weakest real-world adoption of that option.
The interoperability market is fragmenting by design philosophy, not consolidating around a winner. The next 12 months will determine whether this fragmentation is stable — analogous to cloud computing's multi-vendor tiers — or whether one security model achieves sufficient trust, coverage, and compliance to absorb the others. Current data favors the former.