How cross-chain bridge exploits happen

In April 2026, an attacker forged one cross-chain message and walked away with $292M from Kelp DAO’s bridge, the largest DeFi exploit of the year so far. It was one of several. Crystal’s published analysis of stolen crypto withdrawal and transfer patterns (updated May 21, 2026) found that cross-chain bridges are now the dominant withdrawal method for stolen funds, and 2026 has produced a run of further bridge exploits, from that headline loss down to seven- and eight-figure hits on smaller protocols. The pattern repeats because a bridge concentrates trust in one place: whatever process verifies that funds locked on one chain are genuine before releasing them on another. When that verification breaks, a chain-by-chain analytics tool loses the thread at the moment funds cross the boundary, and a threat detection team is left reconstructing the rest by hand.

This piece walks through why bridges keep getting exploited, what happened inside 2026’s largest case, and what it takes to keep following the funds once they cross a chain.

Key takeaways

• In April 2026, an attacker forged a single cross-chain message to steal $292M in restaked ether from Kelp DAO’s bridge, the largest DeFi exploit of the year to date.

• The root cause was structural: the bridge relied on a single verifier, so one forged attestation was enough to authorize the transfer.

• Kelp DAO was one of several bridge exploits in 2026. AFX Trade lost $24.15M in July and Taiko about $1.7M in June, and Crystal’s published analysis of stolen-fund movement shows cross-chain bridges are now the dominant route for moving stolen crypto.

• Once funds cross a bridge, a chain-by-chain analytics tool loses continuity where it matters most, and the rest of the trace becomes a manual, after-the-fact exercise.

• Keeping the trace alive means following the bridge hop inside the same investigation: Crystal Expert’s investigation graph follows a bridge transaction across chains, LayerZero included, so the trace continues on the destination chain and the case file carries every step.

Why are cross-chain bridges such an attractive target?

A cross-chain bridge moves value from one blockchain to another, say from Ethereum to Solana, even though the two networks cannot read each other’s ledgers. Strictly, a bridge does not move a token at all. It locks or burns the token on the source chain and mints or releases an equivalent on the destination chain, based on a message confirming that the first step happened. That confirmation step is usually where a bridge is weakest, more so than the smart contract holding the funds: no chain can check for itself what happened on another, so the whole arrangement is only as trustworthy as the process that reconciles the two sides.

That single point of trust is also where the money sits. Bridges pool large amounts of liquidity to make transfers fast, which turns the verification layer into a high-value target. The same weakness shows up across chains and protocols. In July 2026, AFX Trade lost $24.15M when an attacker obtained enough of its bridge’s validator signing keys to meet the quorum required to authorize a withdrawal: the code worked as designed, and enough compromised keys signed off on the transfer. In June 2026, Taiko lost about $1.7M when a signing key meant to stay sealed inside secure hardware was left exposed on GitHub, letting an attacker forge the cross-chain proofs the bridge relied on. The dollar figures vary by an order of magnitude, but the underlying failure is the same each time: a verification process that a single compromised party, key, or node can satisfy on its own.

What actually happened in the Kelp DAO bridge exploit?

On April 18, 2026, an attacker drained 116,500 rsETH, restaked ether worth approximately $292M and roughly 18% of the token’s circulating supply, from Kelp DAO’s LayerZero-powered bridge. According to LayerZero Labs’ final incident report (May 18, 2026), the bridge used a single-verifier configuration: one Decentralized Verifier Network (DVN) was sufficient, on its own, to approve a cross-chain transfer. The attacker compromised the RPC infrastructure behind that verifier, patched the running RPC nodes in memory so they returned correct responses to LayerZero’s monitoring tools while feeding tampered data to the verifier, and ran a denial-of-service attack against an external RPC provider so the signing service fell back onto the compromised nodes. Those nodes then produced a valid-looking attestation for a cross-chain message that had no real deposit behind it. The report puts the consequence in one sentence: “Because no second independent DVN was required to attest, the destination contract accepted the single valid attestation and unlocked rsETH.” The bridge had no way to tell the forged message from a genuine one, and released the funds as instructed.

The damage did not stay contained to Kelp DAO. The stolen, unbacked tokens were deposited as collateral on lending markets elsewhere in DeFi and used to borrow real assets against them, a chain reaction that spread the loss well beyond the bridge and forced several other protocols to pause markets while the industry worked out what had happened. In response, LayerZero’s verifier network no longer signs attestations for any channel that has it as the only required verifier, the compromised cloud environment was replaced and redeployed on hardened baselines, every channel it participates in now carries a baseline security configuration, and the company has begun security reviews with its partners.

Why couldn’t a chain-by-chain tool follow the funds?

For a threat detection team, the Kelp DAO case is a clean illustration of where single-chain analytics stops being enough. A tool that reads one chain’s ledger follows a wallet without trouble until the funds reach a bridge. After that, the team reconstructs the rest by hand: which new addresses received the funds, which chain or protocol they moved through next, and whether any of it reached a point, such as an exchange or another bridge, where it could still be acted on. Attackers count on that gap. Crystal’s analysis of stolen-fund movement found that bridge-based transfers now execute in minutes, that a stolen asset on Ethereum can move to Tron, then to Solana, and onward to an exchange in a single session, and that more than 75 bridge protocols are in common use. Funds stolen in a bridge exploit are routinely split across newly created addresses within hours and moved through several more protocols before any single team has finished mapping the first hop. Crystal’s earlier explainer on how to investigate emerging risks related to cross-chain crime describes the same dead end from the investigator’s side.

A better risk score on the original chain changes none of this. The trace has to continue across the bridge as one case, with the counterparty on the destination chain tied to a real entity and every step kept for the review that follows. The next section names what does that in Crystal Expert.

What actually closes this gap?

Better security design closes the front half of this problem. A bridge built with no single point of failure in its verification layer, hardened key management, and independent audits is harder to exploit in the first place. That work belongs to the protocol’s own engineering and security review.

The back half is where decision intelligence comes in. Decision intelligence for digital assets is the discipline of putting verified evidence in front of every decision an organization makes about digital assets. Bridges will keep getting targeted however good the engineering gets, since an attacker only needs to find one weak point, so the more useful question for a threat detection team is how fast it can see a bridge exploit, trace where the funds went, and act once one gets through.

Crystal Expert is built for that work. Monitor lets a team keep watch on the source chain and the destination chain alike, so a bridge exploit is visible where the funds land, while there is still time to act. Investigate, the investigation graph in Crystal Expert, lets an analyst follow the trail across the bridge itself, so funds leaving the source chain become the next step in the case; that includes transfers built on LayerZero, the protocol behind the Kelp DAO exploit. From the evidence gathered in Crystal Expert, the team compiles a case report with the audit trail behind every claim open to inspection, ready for the second look a bridge exploit almost always gets.

Ask Crystal, Crystal Expert’s on-demand AI summary, consolidates a transfer’s details, connections analysis, triggered alerts, and prior counterparty history into one narrative, produced in seconds, so an analyst reaches the evidence faster. The judgment call remains theirs. The full comparison with blockchain analytics is in Decision intelligence for threat detection vs. blockchain analytics.

Frequently asked questions

What is a cross-chain bridge exploit?

A cross-chain bridge exploit is an attack on the process that verifies funds locked on one blockchain before an equivalent amount is released on another. The attacker typically forges or manipulates that verification step so the bridge releases funds with no genuine deposit behind them, while the destination chain itself is left intact.

Why do cross-chain bridges keep getting exploited?

Because most bridges concentrate trust in a single verification layer to keep transfers fast, and that layer pools large amounts of liquidity behind it. When a single key, node, or signer is enough to authorize a transfer, one compromise is enough to drain the bridge.

What happened in the Kelp DAO bridge exploit?

On April 18, 2026, an attacker compromised the RPC infrastructure behind Kelp DAO’s single-verifier bridge configuration, forged a cross-chain message, and drained $292M in restaked ether, about 18% of the token’s circulating supply, the largest DeFi exploit of the year to date.

Can stolen funds from a bridge exploit be traced across chains?

Yes, when the tool follows the bridge hop. Once funds cross a bridge and split across new addresses and protocols, you need cross-chain visibility and attribution tied to real entities from the first hop; a manual reconstruction across chains after the fact is slow, and it usually arrives after the window to act has closed.

How does decision intelligence help close the cross-chain visibility gap?

By following the funds across the bridge itself. In Crystal Expert’s investigation graph the bridge hop becomes one more step in the same case, the counterparties on the far side are attributed to real entities, and a full audit trail sits behind the result. That turns a cross-chain flag into evidence you can act on and defend.

What a threat detection team needs before the next bridge exploit

Cross-chain bridges will keep being a target for the same reason they exist: they concentrate liquidity and trust in one place to make transfers fast. The Kelp DAO exploit is the largest 2026 example of what happens when that trust is misplaced, and it won’t be the last. What matters for a threat detection team is being ready before the next one lands: having the evidence to act fast, and to defend that action later.