Bybit Hack: The $1.46 Billion Theft of February 21, 2025

On February 21, 2025, attackers stole roughly 401,346 ETH and related staked tokens — worth about $1.46 billion — from a Bybit cold wallet, making it the largest theft in cryptocurrency history. The attackers did not break Bybit's systems. They compromised a developer machine at Safe, the multisig wallet provider, and used it to show Bybit's signers a transaction that was not the one they were approving.

1. February 21, 2025: The Largest Theft in Crypto History

• Event Background: Bybit, headquartered in Dubai, was among the world's largest cryptocurrency exchanges, holding roughly $16 billion in total assets. Like most large exchanges, it kept the bulk of customer funds in cold storage controlled by a multisignature wallet, moving funds to a warm wallet periodically to service withdrawals.
• Event Details: At approximately 12:30 UTC on February 21, 2025, Bybit detected unauthorized activity during what was a routine scheduled transfer of ether from its Ethereum multisig cold wallet to a warm wallet. CEO Ben Zhou said in a livestream that around 401,346 ETH had been taken, along with staked ether tokens including stETH and mETH. Contemporaneous valuations ranged from $1.4 billion in early reports to $1.46 billion from Elliptic and approximately $1.5 billion in the FBI's public statement; the variation reflects the ether price moving during and after the theft rather than disagreement about the quantity of tokens.
• Technical Architecture Innovation:
  • The scale in context: The previous records were the Ronin Bridge hack in March 2022 at roughly $625 million and Poly Network in August 2021 at about $611 million. Bybit more than doubled the record in a single transaction sequence.
  • Cold storage was not the weak point: The private keys were never extracted. Every signature was produced by a legitimate signer on a legitimate device. The attack targeted what those signers saw.
• Direct Impact: Ether fell several percent on the news, and Bybit faced roughly 350,000 withdrawal requests in the following hours.
• Long-term Significance:
  • A shift in attacker strategy: The largest thefts have moved from exploiting smart contract bugs to compromising the humans and interfaces that authorize transactions.
  • Multisig is not a complete defense: Requiring several signatures protects against one compromised key. It does not protect against all signers being shown the same false information.
• Lessons Learned: An organization's security perimeter includes every vendor whose code renders its transaction approval screens.
• Subsequent Development: The incident is recorded in this site's timeline dataset as the bybit-hack event of February 21, 2025.

2. The Attack Chain: A Compromised Developer Machine

• Event Background: Bybit used Safe, the widely deployed open-source multisignature wallet platform formerly known as Gnosis Safe, which secures a very large share of institutional and DAO treasuries.
• Event Details: Forensic reviews by Sygnia and Verichains, commissioned by Bybit and published within days, traced the attack to infrastructure operated by Safe rather than Bybit. The Safe Ecosystem Foundation confirmed that a Safe developer machine had been compromised, giving the attackers the ability to propose a disguised malicious transaction to an account operated by Bybit. Investigators found that malicious JavaScript served from Safe's AWS S3 bucket — the code that rendered the wallet's web interface — had been modified two days before the attack, on February 19, and was written to activate only for Bybit's specific cold wallet address. Sygnia's review of Bybit's own infrastructure found no evidence of compromise there.
• Technical Architecture Innovation:
  • A targeted supply-chain attack: The malicious code lay dormant for other users of the same interface. Conditioning execution on a single contract address made the compromise nearly impossible to detect through ordinary use, and limited the evidence available before the theft.
  • The trust boundary nobody audits: Organizations audit the smart contracts they interact with and the keys they hold. The web front end that displays what a transaction will do sits between those two, is loaded fresh from a server on every use, and is rarely treated as part of the security model.
• Direct Impact: Safe took its interface offline, rebuilt its infrastructure, rotated credentials, and added new integrity checks before restoring service.
• Long-term Significance:
  • Front-end integrity became a first-class concern: The incident accelerated adoption of independent transaction verification, hardware-wallet clear signing, and self-hosted interfaces among institutional users.
  • Third-party risk in a composable ecosystem: Bybit's own controls were sound and it was still robbed, because the security of a custody operation is the security of its weakest software dependency.
• Lessons Learned: If you cannot independently verify what you are signing, your multisig is a single point of failure wearing a costume.
• Subsequent Development: Hardware wallet vendors and custody providers made independent transaction decoding a headline feature through 2025, and the phrase "blind signing" entered mainstream institutional risk discussions.

3. What the Signers Actually Approved

• Event Background: The mechanical detail matters, because it explains how three legitimate signers approved a transaction that handed over a wallet.
• Event Details: Bybit's signers reviewed what appeared to be a routine transfer to the warm wallet. The interface displayed the correct destination and amount. The transaction actually being signed replaced the logic governing the Safe wallet, giving the attacker control over the contract holding the funds. Once the signatures were collected and the transaction executed, the attacker moved over 400,000 ETH and staked ether tokens to addresses under their control within minutes.
• Technical Architecture Innovation:
  • Delegatecall and upgradeable logic: A Safe wallet's behavior is defined by an implementation contract it delegates to. A transaction that changes that pointer changes everything about how the wallet works, while looking, in raw form, like an ordinary contract interaction. This is why the payload had to be masked rather than hidden — the signers needed to see something plausible.
  • Blind signing at institutional scale: Hardware wallets display transaction data, but complex contract calls often render as unreadable hex. Signers routinely rely on the interface's human-readable summary, which is exactly what the attackers controlled.
  • The defense that would have worked: Independently decoding the transaction on a separate device, or comparing the hash shown by the hardware wallet against one computed offline, would have exposed the mismatch. Several custody providers had recommended this practice before the incident; few enforced it.
• Direct Impact: The theft required no cryptographic break, no software vulnerability in Safe's contracts, and no failure of Bybit's key management. It required only that the screen lie.
• Long-term Significance:
  • A new standard threat model: Post-Bybit, institutional custody designs assume the signing interface may be hostile and require out-of-band verification.
• Lessons Learned: Signing is a promise about a specific set of bytes. Any workflow where a human approves a summary instead of the bytes has a gap an attacker can occupy.
• Subsequent Development: The incident became the standard case study in institutional custody training, and clear-signing standards that render transaction intent on the hardware device itself gained rapid adoption afterward.

4. Bybit's Response and the Laundering Race

• Event Background: Exchanges that suffer losses of this magnitude usually halt withdrawals, which typically triggers the run that finishes them. Bybit did the opposite.
• Event Details: Ben Zhou stated publicly within hours that Bybit remained solvent and could cover the loss even if nothing was recovered. Withdrawals stayed open and roughly 350,000 requests were processed. Bybit closed the gap in its ether reserves within about 72 hours through a combination of bridge loans, large deposits from counterparties, and open-market purchases, then published proof-of-reserves data. It also launched a bounty program offering a share of recovered funds to anyone who helped freeze or trace the stolen assets. Meanwhile the attackers moved quickly: analysts at Elliptic tracked roughly a billion dollars of the stolen ether being converted and dispersed within 48 hours, with a large share routed through the cross-chain protocol THORChain into bitcoin, then split across thousands of addresses on multiple blockchains.
• Technical Architecture Innovation:
  • Speed as a laundering strategy: The attackers assumed that freezing and coordination take time, so they converted through decentralized, non-custodial venues before centralized exchanges could react. The volume pushed THORChain to record activity levels and prompted a public debate about whether the protocol should have intervened.
  • Transparency as a countermeasure: Because every movement was on chain, analytics firms published address lists in near real time and the FBI released 51 Ethereum addresses linked to the theft, asking node operators, exchanges, bridges, and DeFi services to block related transactions.
  • Why open withdrawals were the right call: Solvency is demonstrated by processing withdrawals, not by announcing it. Keeping the gates open converted a potential FTX-style collapse into a large but survivable loss.
• Direct Impact: Bybit continued operating without interruption, an outcome that distinguishes this incident from nearly every other theft of comparable relative size.
• Long-term Significance:
  • A playbook for exchange incident response: Immediate disclosure, open withdrawals, external forensics published quickly, and reserve replacement within days is now the reference standard.
  • The limits of tracing: Funds were traced extensively and frozen only partially. Visibility on a public ledger does not equal recoverability once assets pass through decentralized swap venues.
• Lessons Learned: In a custody crisis, credibility is produced by letting customers leave. Bybit is the clearest demonstration that a solvent exchange can survive a record theft.
• Subsequent Development: The event intensified scrutiny of cross-chain swap protocols and revived policy discussions about whether decentralized bridges should implement compliance controls.

5. Attribution and Common Misconceptions

• Event Background: Attribution came quickly and from multiple independent directions, which is unusual.
• Event Details: On February 26, 2025, the FBI issued a public service announcement stating that North Korea was responsible for the theft of approximately $1.5 billion in virtual assets from Bybit on or about February 21, and identifying the activity as "TraderTraitor," a cluster also tracked as Lazarus Group and APT38. Independent investigator ZachXBT had linked the addresses to previous DPRK-attributed thefts within a day, citing overlap with the Phemex, BingX, and Poloniex incidents. The FBI's announcement included 51 Ethereum addresses holding or having held the stolen assets.
• Technical Architecture Innovation:
  • State-sponsored theft as a funding model: North Korean groups have stolen billions in cryptocurrency across recent years, and the Bybit theft alone represents a substantial share of a single year's total. The targets and methods are chosen for scale, and the operational patience — a developer machine compromised days ahead, code that activates for one address — reflects a resourced program rather than opportunistic crime.
  • Bybit was not hacked, precisely speaking: The most common misstatement is that Bybit's security failed. Its infrastructure was not compromised; its wallet provider's was. The distinction matters for anyone drawing lessons about exchange security.
  • This was not a smart contract exploit: No bug in Safe's contracts was used, and no flaw in Ethereum was involved. Describing it as a "DeFi hack" or a "smart contract exploit," as some coverage did, misidentifies the failure.
• Direct Impact: The theft prompted regulators in multiple jurisdictions to examine custody and vendor-management requirements for exchanges, and drove sanctions attention toward the laundering infrastructure used.
• Long-term Significance:
  • The record it set: At roughly $1.46 billion, the Bybit theft is more than twice the size of any previous single incident and larger than most annual industry loss totals.
  • The lesson the industry actually took: Not "cold storage is unsafe," but "verify transactions independently of the interface that constructs them."
• Lessons Learned: As protocol-level security has improved, attackers have moved to the seams between organizations — vendors, interfaces, and the humans reading them. That is where the next record theft is most likely to originate.
• Subsequent Development: Bybit recovered, continued to publish reserve attestations, and the bounty program continued paying out to analysts who traced portions of the stolen funds through subsequent laundering hops.