Hyperliquid Bridge

Rejected venues wait the longest for re-review; a rejection has to earn another look before the scheduled date.

REJECTED ON A STANDING TRANSPARENCY FACT, not a temporary evidence gap. The bridge locks USDC on Arbitrum and mints it 1:1 as collateral on HyperCore, Hyperliquid’s own custom consensus chain, secured by the same roughly two-thirds-of-stake validator quorum that secures the exchange itself — not an Ethereum-native fraud-proof or ZK system, and not economically independent of the exchange it serves. Validator node software has been closed-source since early 2025: validators run a binary they cannot independently audit. That is a standing characteristic of the system, not a fact this review expects to change soon, and it blocks certifying what is actually running regardless of how the validator set’s decentralization trends. Hyperliquid’s own documentation states the tracked legacy Arbitrum bridge now holds less than 10% of USDC supply on HyperCore, meaning the surveyed TVL figure may already track a shrinking minority path.

Kill criteria: the events that revoke this verdict
• Validator node software becomes open-source or is independently, reproducibly verified against deployed binaries
• Foundation-affiliated validator stake exceeding 50% of total is reversed and stays below that level for 90 days
• USDC-specific bridge withdrawal timing and dispute mechanics are independently confirmed and pass a proposed-size stressed-exit test
• A new independent audit is published covering current bridge code, superseding the 2023 Zellic reports
• Any confirmed exploit of the bridge contract itself, as distinct from an exchange market-mechanic event

The research file

Mechanism

USDC locked in an Arbitrum contract is minted 1:1 as the collateral asset on HyperCore. This is lock-and-mint, not a rollup with Ethereum-native fraud or validity proofs: validation of deposits and withdrawals relies on Hyperliquid’s own validator set signing off under the same HyperBFT consensus that orders the exchange’s own trades. Hyperliquid’s own documentation states the tracked legacy Arbitrum bridge now carries less than 10% of USDC supply on HyperCore, which means most current USDC enters through a different, unreconciled path this review did not verify.

Control and governance

Twenty-seven validators as of June 2026, up from four at launch. Foundation-run nodes hold about 49.3% of staked HYPE; 22 independent operators hold the rest, down from roughly 81% foundation concentration in January 2025 — real, if recent, decentralization. Validator entry requires staking more than 1M HYPE; validators can be jailed by peer vote, but foundation-affiliated stake is large enough to materially influence that vote. Validator node software has been closed-source since early 2025, so no outside party can independently confirm what code a validator is actually running. The bridge contract and L1 staking logic were audited by Zellic, most recently in November 2023 — audits now roughly three years old against a codebase that has changed materially since.

Incident record

No verified exploit of the bridge contract itself was found. Two frequently cited incidents belong to the exchange’s market mechanics, not the bridge, and should not be conflated with it: the March 2025 JELLY token manipulation, where validators voted to intervene and settle a manipulated position — demonstrating real emergency intervention capability at the L1 level, for better or worse — and the November 2025 POPCAT manipulation, a roughly $4.9M market-mechanic loss. A widely indexed $1.2B forged-mint incident belongs to Hyperbridge, an unrelated Polkadot interoperability protocol, and does not apply here; it is noted only to prevent a citation error.

Exit

HYPE unstaking from the staking module to spot takes seven days through a queue capped at five pending withdrawals per address, but that governs HYPE, not the USDC this bridge carries. This review could not confirm the specific USDC bridge-withdrawal timing or dispute mechanics from Hyperliquid’s public documentation; the gitbook bridge page was unreachable, leaving only the GitHub bridge code and the 2023 Zellic audit as technical sources. An advisor cannot certify a stressed exit path that has not been independently confirmed.

Comparison

A standard rollup bridge secured by Ethereum-native fraud or validity proofs inherits Ethereum’s security for the dispute process itself, independent of the rollup operator’s good faith. This bridge’s security is only as strong as Hyperliquid’s own 27-validator set and its closed-source client — a single-application trust model, not a general-purpose, independently verifiable one. The tradeoff for bridging into an app-specific chain at all is direct exposure to one exchange’s validator and governance risk in return for access to its liquidity; this review does not find that trade currently certifiable.

Sources

The claims above trace to these. Where a number could not be independently verified, the thesis says so.

• Halborn — explained: the Hyperliquid hack, March 2025 · secondary · accessed 2026-08-17
  Supports: JELLY incident, validator intervention
• Halborn — explained: the Hyperliquid hack, November 2025 · secondary · accessed 2026-08-17
  Supports: POPCAT incident, loss figure
• crypto.news — who actually runs Hyperliquid: the governance audit · secondary · accessed 2026-08-17
  Supports: validator set analysis, foundation concentration
• Hyperliquid Docs — audits · primary · accessed 2026-08-17
  Supports: Zellic audit dates, scope
• Hyperliquid Docs — staking · primary · accessed 2026-08-17
  Supports: unstaking queue, seven-day timing
• Hyperliquid Docs — full export (llms-full.txt) · primary · accessed 2026-08-17
  Supports: bridge design references, less than 10% legacy Arbitrum share
• OneKey — Hyperliquid validator risk, 2026 briefing · secondary · accessed 2026-08-17
  Supports: stake distribution, decentralization trend

Inherited controls

The verdict above grades the protocol layer. Every position also inherits the asset it holds and the chain it settles on. The least safe layer sets the position’s grade, and the position table names which one that is.