Polygon co-founder Sandeep: Writing after the Chain Bridge explosion

Author: Sandeep

Compiled by: Carnival, ChainCatcher

This weekend has been heartbreaking. Three security incidents have occurred on the cross-chain bridge in just three weeks. These days, I haven't focused on the specific details of any of the attacks, but rather on studying the patterns underlying all these incidents.

On April 1, Drift reported a loss of $285 million.

On April 13, Polkadot Hyperbridge minted 1 billion unsupported tokens through a replay proof; the losses would have been even greater if the target chain's on-chain liquidity had not been already very low.

On April 18, KelpDAO lost $292 million. Prior to this, projects such as Wormhole, Ronin, Harmony, BNB Bridge, Nomad, and Multichain also suffered similar fates.

First and foremost, I want to express my deepest respect to all the teams that worked tirelessly and proactively during this stressful weekend. I have absolutely no intention of holding anyone back while they were handling the emergency.

We've all been in similar situations, and the team currently releasing patches is working hard. Kelp's emergency suspension of the multi-signature mechanism prevented two subsequent asset theft attempts; otherwise, the losses would have increased by another $200 million.

I want to emphasize that what happened this weekend isn't just a Kelp issue, but a design choice the entire industry has been making all along. Currently, most cryptocurrency cross-chain infrastructure still operates like a notary public.

Whether you call it DVN, relay set, oracle committee, or multi-signature, it is essentially a small committee that monitors activity on one chain and certifies it on another chain.

If the committee or the price information it relies on is compromised, this notary will not hesitate to endorse the lie. The name of the agreement may change, but the premise of trust remains unchanged.

@moo9000 gave it a very fitting name: MultisigFi.

This description is very apt. Regardless of what you call this underlying committee, its trust model is the same, and what has happened over the past three weeks has painfully revealed how this model collapses as it scales up.

A recent scan of LayerZero's active applications by Dune revealed that 47% of applications run on a 1/1 validator configuration, 45% run on a 2/2 configuration, and less than 5% use a more robust security configuration.

This means that for 90% of cross-chain applications in the current production environment, one or two compromised signers are the only security barrier between user funds and attackers.

Five years ago, this might have been considered a reasonable security setup. At that time, cross-chain bridges were handling only a few million dollars in transactions, and no one was conducting large-scale probing.

But this is absurd in 2026. The same architecture can now move billions of dollars! Furthermore, AI-assisted tools are constantly discovering runtime configuration vulnerabilities at machine speed. The attack surface is growing exponentially, while security models are stagnating.

To clarify, this article is not intended to pit Polygon against other companies. Years ago, we also built a similar trust assumption into our products in earlier versions. We learned from that experience, and the entire industry has benefited greatly from it.

In the process, some of us continued to build using the committee model, while others staked the entire company on ZK (zero-knowledge proofs).

Our commitment to zero-knowledge proofs is not just talk: In July 2024, we launched zero-knowledge proofs for the Agglayer bridge, which has been online for over a year and is settling cross-chain transactions on a large scale every day. Frankly, what happened this weekend has further solidified my confidence in this.

Zero-knowledge proofs replace the work previously done by committees. They are like miniature encrypted receipts that prove the computation was indeed performed correctly, and any machine on Earth can verify it in milliseconds.

Either the mathematical verification is successful and the transfer is completed, or the mathematical verification fails and the assets remain unchanged. No operator can be bribed, no RPC can be tampered with, no quorum needs to be coordinated, and no one will be sitting in a room at 3 a.m. on a Saturday to decide whether your money is safe.

In addition, we have what is called "pessimistic proof." The simplest way to understand it is: don't trust anyone's on-chain ledger.

Each chain connected to Agglayer has a dynamic ledger that records the receipt and sending of assets, and accounts must remain balanced for any withdrawal to be finally confirmed. Regardless of the reason or whether upstream messages are forged, a chain can never withdraw more assets than recorded.

Mathematical rules do not allow such situations to occur. Agglayer enforces this rule through a Succinct SP1 proof system built on Polygon Plonky3.

If the situation from last weekend had occurred on Agglayer, the pessimistic proof would have immediately blocked withdrawals because there was no deposit record, so the funds would absolutely not have been transferred.

The same accounting mechanism can detect the infinite issuance vulnerability in Wormhole and BNB Bridge, and the replay proof vulnerability in Hyperbridge.

These vulnerabilities are fundamentally different, but they all boil down to the same problem: cross-chain bridges release assets that are not supported on the other end. Agglayer can prevent all of these situations from happening before settlement occurs.

This is not just theory. Despite most DeFi transactions being suspended this weekend, Agglayer still processed approximately $200 million in bridge transactions, completely unaffected.

Katana's native integration with Agglayer maintained zero risk exposure throughout the incident. Our security team suspended LayerZero integration within the Polygon ecosystem prior to the public disclosure of the root cause, and our product and support teams remained in communication with institutional partners throughout the weekend.

After nearly six years of development, the Polygon platform has completed $24 trillion in settlements, processed 7 billion transactions, and boasts a 99.99% uptime. The Agglayer platform also has zero cross-chain bridging vulnerabilities. This is why we spent years building Agglayer: security has always been our top priority.

I'm listing these numbers not to show off, but because to confidently walk into an institution and tell them that cryptocurrency is ready to handle a large volume of payment transactions, you have to demonstrate these tangible achievements.

It's cheaper and faster to build cross-chain bridges based on the fact that committees do so; I understand why the team wanted to build them, and we also built earlier versions. However, the capabilities of attackers have indeed changed.

The Lazarus group has been attacking these designs since 2022, and there are no signs of it slowing down. AI-assisted audits can now uncover all the configuration errors previously hidden beneath complex hierarchical structures. These attacks will not disappear. Mathematics will eventually compensate for the committee's shortcomings.

Over the past two or three years, this industry has processed trillions of dollars in transactions annually. We are demanding that banks and payment companies invest huge sums of money in payment systems that still rely on one or two signatories making the right judgment on Saturday nights. That's our demand; you'll realize how absurd it is once you hear it.

We must do better, and we already know how.

Nevertheless, we must acknowledge that LayerZero is currently disabling 1/1 settings (single signatures) across the industry. This is the right decision, as it will significantly enhance cross-chain security; I fully support this measure. Other teams will also continue to refine their committee designs. This work is crucial.

But the bigger shift lies in the architecture. Zero-knowledge proofs are indestructible, immune to social engineering attacks, and unaffected by weekends. Mathematics either holds true or it doesn't; if it doesn't, everything comes to a standstill.

This is the direction the industry is heading, and the pace of development is even faster than it was a month ago, which is good news for every builder and every organization entering the industry.

This week, all teams building cross-chain infrastructure should ask themselves: Do I really need a committee? Strengthening existing committees is merely a last resort.

Agglayer is open source, with no protocol fees and no access restrictions. Any team preparing to transition from trusted proof-of-trust mechanisms to cryptographic verification can access it. If you are currently running a cross-chain bridge and the events of the past three weeks have made you rethink your trust model, please contact us.

This is not a competitive moat we've built up; this is infrastructure that the entire industry should use.

The fate of cryptocurrencies over the next decade will depend on teams willing to start building more complex architectures now. Cryptographic proofs are harder to establish than notarizations, but they won't collapse on the weekend, and they can scale to the trillions of dollars that cryptocurrencies currently need to handle.

Would you prefer a committee ruling or a mathematical proof? We chose the latter. I hope more people will make the same choice.

After this weekend, I'm even more convinced of the importance of zero-knowledge cross-chain technology. In challenging times, a clear architecture is crucial.