Cross-chain bridges have produced $329 million in exploit losses through May 2026, accounting for the two largest DeFi hacks of the year. The KelpDAO LayerZero bridge lost $292 million on April 18. Drift Protocol lost $285 million on April 1. Both attacks targeted off-chain infrastructure — compr...
"Most bridges don't fully verify what happened on another chain. Instead, they rely on a smaller system to report it. That system becomes the thing you trust." — Ben Fisch, CEO of Espresso Systems
Cross-chain bridges have produced $329 million in exploit losses through May 2026, accounting for the two largest DeFi hacks of the year. The KelpDAO LayerZero bridge lost $292 million on April 18. Drift Protocol lost $285 million on April 1. Both attacks targeted off-chain infrastructure — compromised RPC nodes and social-engineered multisig signers — rather than smart contract bugs. The failure modes are structural: validator-set bridges become economically insecure when the cost to corrupt a validator threshold falls below the value of assets the bridge secures.
Cumulative bridge losses since 2022 now exceed $2.8 billion, representing roughly 40% of all value hacked in Web3 over that period. Bridge TVL dropped below $45 billion in late June 2026, down from approximately $50 billion in May, a 10% decline driven by exploit-triggered withdrawals. Meanwhile, intent-based bridge architectures — which eliminate custodial pools entirely — have recorded zero protocol-level exploits to date, processing over $20 billion in cumulative volume. The data suggests a measurable divergence in security outcomes between architectural models.
Eight major bridge incidents have been recorded through May 2026, according to PeckShield. The combined damage: $328.6 million. Two incidents account for 88% of that total.
| Date | Protocol | Loss | Attack Vector | |------|----------|------|---------------| | Feb 2, 2026 | CrossCurve | $3M | Spoofed Axelar messages; missing access control | | Apr 1, 2026 | Drift Protocol | $285M | Social-engineered multisig; fake oracle collateral | | Apr 18, 2026 | KelpDAO | $292M | Compromised RPC nodes; forged LayerZero DVN message | | May 18, 2026 | Verus-Ethereum | $11.5M | Forged Merkle proof; missing field validation |
Additional smaller incidents brought the running total to $328.6 million by end of May. In that month alone, bridges accounted for $28.6 million of the approximately $70 million in total crypto exploit losses — a 42% share from a single protocol category.
Q2 2026 was the most-hacked quarter on record across all of DeFi, with roughly $746 million drained across approximately 70 exploits. Bridge-related losses constitute a material share of that figure.
On April 18, attackers linked to North Korea's Lazarus Group drained 116,500 rsETH — approximately 18% of the token's circulating supply — from KelpDAO's LayerZero bridge in under 46 minutes, according to analysis by Chainalysis and TRM Labs.
The attack chain:
The root cause, according to LayerZero, was KelpDAO's use of a 1-of-1 DVN configuration — a single verifier — which "directly contradicted" LayerZero's standing recommendation for diversified multi-DVN setups. The code functioned as designed; the trust assumption was flawed.
The Arbitrum Security Council froze over 30,000 ETH of the attacker's downstream funds. KelpDAO paused contracts to block a second $95 million theft.
As Ben Fisch of Espresso Systems noted: "Attackers compromised nodes and fed the system a false version of reality. The bridge worked as designed. It just believed the wrong information."
On April 1, approximately $285 million was drained from Drift Protocol, Solana's largest decentralized derivatives platform, in a 12-minute window, according to reporting by The Hacker News and TRM Labs.
The attack was attributed with medium confidence to UNC4736, a North Korean state-sponsored group also tracked as AppleJeus, Citrine Sleet, Golden Chollima, and Gleaming Pisces. It was not a code exploit. The campaign began in fall 2025.
Attack sequence:
Both attacks share a pattern: neither exploited a smart contract bug. Both targeted human and infrastructure layers that sat outside the auditable code perimeter.
Bridge exploits have produced the following estimated losses by year, compiled from CertiK, PeckShield, Phemex, and Chainalysis data:
| Year | Estimated Bridge Losses | Notable Incidents | |------|------------------------|-------------------| | 2022 | ~$1.9B | Ronin ($625M), Wormhole ($320M), Nomad ($190M) | | 2023 | ~$250M | Multiple smaller incidents | | 2024 | ~$1.19B | Bridges/cross-chain messaging <5% of protocols by count | | 2025 | Data incomplete | — | | 2026 (through May) | $328.6M | KelpDAO ($292M), Drift ($285M) | | Cumulative | >$2.8B | ~40% of all Web3 hacks |
The ratio is persistent: bridges represent a small fraction of DeFi protocols by count but generate a disproportionate share of total dollar losses. In 2024, bridges and cross-chain messaging protocols produced $1.19 billion in losses despite representing fewer than 5% of monitored protocols, according to data compiled by Yellow Research.
Bridge TVL stood at approximately $50 billion in May 2026 before dropping below $45 billion by late June — a 10% decline attributed to exploit-driven capital flight, according to Times of Blockchain.
The bridge category is not monolithic. Three distinct architectural models produce measurably different security outcomes.
The traditional model: users lock assets in a contract on the source chain; the bridge mints wrapped tokens on the destination chain. A custodial pool backs all wrapped tokens. If the pool is drained, wrapped tokens on all connected chains lose backing simultaneously.
KelpDAO, Wormhole (2022), and Ronin (2022) all used variants of this model. The structural risk is that a single breach can drain the entire reserve backing tokens across 20+ chains.
Bridges secured by a committee of validators who attest to cross-chain state. The security model breaks when the cost to corrupt a validator threshold — through bribery, social engineering, or key compromise — falls below the value of bridged assets.
As Sergej Kunz, co-founder of 1inch, stated: "As long as we rely on validator-based bridges, these problems will continue."
Intent-based bridges (Across Protocol, deBridge, Relay) eliminate custodial pools entirely. Users sign a desired outcome. Professional solvers compete to fill the order using their own capital. Settlement is verified on-chain.
deBridge has processed over $20 billion in volume across 30+ security audits with zero protocol-level exploits, according to its published documentation. Across Protocol operates similarly with no recorded major exploit.
The structural advantage: there is no shared pool to drain. Solver capital is individually held. A compromise of one solver does not cascade to others.
The limitation: intent-based designs require active solver markets. For illiquid pairs or large orders, execution may be slower or more expensive.
Less than 2% of DeFi's total value locked carries insurance coverage, according to CoinDesk reporting from May 2026. The gap is most acute for bridge users.
Nexus Mutual, the largest on-chain insurance protocol, has paid out more than $18.5 million across all historical claims — a figure dwarfed by the $292 million lost in the KelpDAO incident alone. Notable prior payouts include $5.09 million for the TribeDAO/Rari Capital hack (April 2022) and $2.39 million for the Euler Finance exploit.
Recovery rates after bridge exploits remain low. In May 2026, only approximately $9.4 million of $68.3 million in total monthly crypto losses was returned, according to PeckShield data. Negotiations, bounties, and law-enforcement pressure occasionally produce results — the Arbitrum Security Council's freeze of 30,000 ETH after KelpDAO is one example — but users should not assume recovery is probable.
The structural challenge: insurance pools often share the same smart contract and oracle dependencies as the protocols they cover. An exploit that drains a bridge may simultaneously impair the insurance pool's ability to pay claims.
The data from 2022 through mid-2026 establishes a clear pattern: custodial bridge architectures concentrate risk in ways that produce catastrophic, single-point-of-failure losses. The attack surface has shifted from smart contract vulnerabilities — which audits can partially address — to infrastructure and human-layer compromises that sit outside the auditable perimeter.
Intent-based architectures offer a structural alternative by eliminating shared custodial pools, though they introduce different tradeoffs in liquidity depth and execution guarantees. The zero-exploit record across Across, deBridge, and Relay — while not proof of invulnerability — reflects the security advantage of removing the honeypot.
The insurance market has not scaled to meet the risk. At less than 2% coverage and cumulative payouts smaller than a single mid-tier exploit, on-chain insurance functions as supplementary protection rather than systemic backstop.
For protocols evaluating cross-chain infrastructure, the economic calculation is straightforward: bridge architecture is a risk management decision, not merely a connectivity feature. The $2.8 billion in cumulative losses quantifies the cost of treating it otherwise.