NextFin News - On the night of February 21, 2025, Bybit's chief executive Ben Zhou logged into his computer and approved what looked like a routine transfer of Ethereum from the exchange's cold wallet to its warm wallet. Thirty minutes later his chief financial officer called with a trembling voice: "All of the Ethereum is gone." Within about half an hour, $1.5 billion had vanished - the largest theft in the history of cryptocurrency. It happened not because the cryptography failed, but because the screen Zhou was looking at was lying to him.
The heist, attributed by the Federal Bureau of Investigation to North Korea's TraderTraitor unit, did not crack a vault. It rewrote the user interface of a free piece of software that Bybit trusted to coordinate its most sensitive transactions. The attack exposes a structural weakness at the heart of crypto's security model: the industry built trillion-dollar infrastructure on free, open-source tooling guarded by human approval, and a single compromised developer workstation at a third-party vendor was enough to turn those human approvers into unwitting accomplices.
The Half-Hour Unraveling: What Actually Happened
The numbers set the scale of the failure. Attackers drained Bybit's cold wallet of 401,347 ETH plus 90,375 stETH, 15,000 cmETH and 8,000 mETH - a combined haul the exchange valued at roughly $1.5 billion. Cold wallets are supposed to be the industry's safest storage, with private keys held offline, far from the internet-connected "hot" wallets exchanges use for daily operations. Bybit's cold wallet was protected by a multisignature setup requiring at least three authorized signers to approve any movement of funds. On paper, that is defense in depth. In practice, it created a false sense of security.
The weakness was not in Bybit's wallet at all. It was at Safe{Wallet}, the third-party multisignature platform Bybit used to coordinate approvals. In early February, a Safe developer fell for a social engineering attack and his Mac workstation was compromised. Forensic work later dated the earliest malicious activity to February 4. The next day, attackers used the developer's AWS access token to enter Safe's cloud infrastructure, where they operated undetected for more than two weeks, through February 21.
Here is the mechanism that made the heist possible. Safe's security configuration required multi-factor authentication to be re-confirmed for temporary session credentials every 12 hours. The attackers could not register a new MFA device on the compromised account - that control held. So they did something simpler: they stole an already-active session token from the developer's machine and rode it, carefully scheduling their activity to match the developer's normal working hours so nothing looked unusual.
Once inside, they did not touch the blockchain. They touched the web page. The attackers injected malicious JavaScript into the Safe interface that Bybit employees used to review transactions. The code was precise: it checked whether the person signing was one of Bybit's authorized signers, and if so, it silently swapped the transaction about to be signed. It replaced the destination address with one the attackers controlled, changed the operation type to a "delegatecall" that would hand over control of the cold wallet's smart contract, and set the transaction value displayed as zero.
Then came the sleight of hand. After the signers approved the tampered transaction, the malicious code restored the original transaction data in the response shown on screen. Zhou and his team saw exactly what they expected to see: a legitimate internal transfer, approved, confirmed. The blockchain recorded something entirely different. By the time anyone realized the destination address on the ledger did not match the one on the screen, the funds were already moving through wallets the attackers controlled.
The entire approval-to-discovery window lasted roughly 30 minutes. Multisig had done its job - three people had approved the transaction. It just was not the transaction they thought they were approving.
Why the Industry's Favorite Defense Failed
The Bybit heist is often described as a multisig failure. That is wrong, and the distinction matters. Multisignature security solved one problem: it prevents a single compromised key from draining a wallet. It never solved the problem of a compromised interface showing signers the wrong transaction. The attack exploited exactly that gap - the space between what a human sees and what the network executes.
This is a supply-chain attack in the purest sense. Bybit did everything the security playbooks recommend: offline cold storage, multiple signers, human review. But it had outsourced the coordination layer to a free product used by thousands of other organizations. The attackers did not need to breach Bybit. They needed to breach one developer at one vendor, and Bybit inherited the compromise automatically.
The economics of the target explain the method. North Korean cyber units have spent years learning that crypto firms are softer targets than regulated banks. Chainalysis data shows the pattern clearly: North Korean actors stole $660.5 million across 20 incidents in 2023, then $1.34 billion across 47 incidents in 2024 - a 102.88% increase in value stolen in a single year. The Bybit heist alone exceeded everything they took in all of 2023. A senior Biden administration official estimated in 2024 that roughly half of North Korea's foreign-currency earnings came from cybercrime.
For a state under sanctions, crypto theft is not just profitable - it is strategically efficient. There are no correspondent banks to pressure, no borders to cross, and a global network of permissive exchanges and mixing services ready to convert the proceeds. The FBI's public attribution, issued on February 26 under the code name TraderTraitor, named the group as a subunit of the Reconnaissance General Bureau's 3rd Bureau, North Korea's primary foreign intelligence service.
"Every minute matters for the hackers who are trying to confuse the money trail and they are extremely sophisticated in what they're doing," said Dr. Tom Robinson, co-founder of blockchain investigators Elliptic. "I imagine they have an entire room of people doing this using automated tools and years of experience. We can also see from their activity that they only take a few hours break each day, possibly working in shifts to get the crypto turned into cash."
The laundering operation that followed was industrial in scale. Within the first 48 hours, at least $160 million of the stolen funds had already been laundered. The attackers moved the ether through multiple intermediary wallets, decentralized exchanges and cross-chain bridges, converting it into Bitcoin and eventually through mixers. By March 20, Zhou reported that 86.29% of the stolen ETH had been converted to BTC. The pace was deliberate: this was not a panicked cash-out, but a professionalized conversion pipeline running in shifts around the clock.
The Market Reaction: Contained Panic, Unresolved Trust
The immediate market response was sharp but short-lived. In the week after the breach, Ethereum fell about 18% to a local low near $2,100, and on February 25 alone it dropped 8.46% to $2,414.29, around its lowest level since the previous October. Bitcoin slipped below $95,000, down about 20% from its January record high of $109,071. Bybit itself faced a surge of withdrawal requests - more than 350,000 of them - as users tested whether their funds were actually safe.
What is striking is not the size of the drop, but how quickly it stabilized. Zhou moved fast to contain the run. Bybit's own timeline shows that within 10 hours, 99.994% of those withdrawal requests had been processed, and less than 12 hours after the breach the platform had resumed normal operations. Zhou assured customers that no user funds were lost and announced that Bybit had replenished the stolen ether with loans from investors. The lesson for the market was clear - the exchange survived, but only because it had deep-pocketed backers willing to backstop a loss no ordinary firm could absorb.
The recovery effort has been partial and slow. Chainalysis reported working with industry contacts to freeze more than $40 million of the stolen funds. Bybit launched a bounty program offering up to 10% of recovered amounts, and a group of bounty hunters - including blockchain analyst ZachXBT and protocols Mantle and Paraswap - shared rewards for helping identify and block stolen money. Yet the broader picture is grim: roughly 20% of the funds have "gone dark," meaning they are unlikely ever to be recovered, and more than $90 million was funnelled through a single non-compliant exchange that disputed whether it should have blocked the transfers.
Cyclical or Structural: Why This Will Not Be the Last
Every major crypto heist is followed by the same question: will this one change the industry? The answer requires separating the cyclical from the structural. The cyclical part is real and has largely reverted - the panic selling subsided, Bybit remained solvent, and trading volumes recovered. Confidence in this specific exchange proved resilient because the loss was socialized across investors rather than users.
The structural part will not revert on its own. The vulnerability that enabled the heist - reliance on free, shared infrastructure with a human-in-the-loop approval process that can be deceived at the interface - remains widespread across the industry. Safe{Wallet} is not a niche tool; it is used by thousands of organizations to manage billions in assets. The attack did not exploit a bug in Safe's smart contracts. It exploited the fact that the interface layer, the very thing humans trust when they click "approve," sat outside the security boundary that multisig protects.
This is a regime-level problem, not a one-off bug. Three pieces of evidence support that call. First, the attack surface is permanent: as long as signing happens through a web interface served from infrastructure that can be compromised upstream, the deception vector exists. Second, the economics are permanent: sanctioned states face no better alternatives for raising hard currency, and crypto remains the most efficient channel available. Third, the incentive structure is misaligned - vendors of free security tooling have limited resources for the kind of defensive hardening that nation-state attackers will patiently probe, while their customers assume "multisig" means "safe."
The strongest counter-thesis is that the industry is already adapting. Proponents point to the rapid freezing of more than $40 million through public-private cooperation, the bounty program that turned independent researchers into a distributed surveillance network, and the fact that no user lost money at Bybit. They argue that the ecosystem's transparency - every transaction visible on a public blockchain - makes crypto theft uniquely traceable compared to traditional bank robbery.
That argument mistakes traceability for recoverability. The blockchain is public, but the off-ramps are not. Once funds cross into a permissive exchange, pass through a mixer, or convert into privacy-focused assets, the trail ends in practice even if it remains visible in theory. The $90 million that moved through a single uncooperative exchange illustrates the structural gap: there is no global mechanism to freeze illicit crypto flows, only a patchwork of compliant and non-compliant venues that attackers route between at will.
The falsifying signal for the structural view is specific and observable: if, within 18 months, the majority of large custodians migrate to hardware-signing devices that display and verify transaction details independently of the vendor's web interface - a class of products that already exists - then the interface-deception vector would be closed and the structural thesis would weaken materially. Absent that migration, the same playbook will be reused.
What Comes Next
In the short term, expect continued volatility around any news of further conversions. The attackers still hold the bulk of the stolen Bitcoin, and any large movement could pressure prices. Bybit's bounty program will keep producing incremental freezes, but the recoverable portion is likely capped near current levels.
Over the medium term, the pressure shifts to infrastructure vendors. The Bybit heist makes the business case for independent transaction-verification hardware - devices that show signers the actual destination address and amount on a separate screen, cryptographically verified, before a signature is released. Firms that treat multisig as a complete solution rather than one layer will be the next targets.
In the long term, the structural question is whether regulation catches up to the laundering infrastructure. The United States has moved toward a more crypto-friendly stance, but the Bybit heist demonstrates that friendly regulation without enforceable off-ramp controls simply gives attackers more compliant venues to avoid. The specific signal to watch is whether major jurisdictions mandate real-time screening of deposits against known illicit addresses at exchanges and bridge services - and whether that mandate has teeth against non-compliant venues.
The base case is more heists of the same shape: supply-chain compromises of shared tooling, interface deception, and rapid conversion through permissive venues. The upside case is that this heist becomes the industry's wake-up call, driving adoption of hardware-backed signing and enforceable off-ramp screening. The downside case is a repeat against a less-capitalized firm that cannot backstop the loss with investor loans - and that is where the systemic risk actually lives.
The uncomfortable truth the Bybit heist leaves behind is that the industry spent a decade building stronger vaults while leaving the front door's peephole wide open. The next $1.5 billion will not be stolen by breaking encryption - it will be stolen by showing the right person the wrong screen at the right moment, and waiting for them to click approve.
Explore more exclusive insights at nextfin.ai.
