Gracy Chen said the number out loud. In an interview with BeInCrypto, Bitget's CEO confirmed that attackers drained roughly $387.5 million from the exchange โ the largest single loss it has ever taken. But the figure she would not say carries more weight. The intrusion came through "a previously unknown vulnerability in a third-party product," she explained. That vendor still has no name.
For anyone who has stood inside an exchange's operations room during a crisis, that omission is the loudest sentence in the whole story. In 2022, I ran "Transparency Tuesdays" for a mid-tier exchange โ opening our cold wallets live on camera while 50,000 traders watched, waiting to see whether their funds still existed. I learned something that no audit report ever taught me: the breach you can see is survivable. The dependency you never mapped is the one that ends you.
To understand why this event matters, you have to place it on a timeline. February 2025 brought the Bybit compromise โ roughly $1.5 billion, the largest in crypto history. September brought Bitget, with a loss roughly a quarter of that size. Two major centralized exchanges, two similar attack patterns, seven months apart. That is not coincidence. It is a pattern. The parallel is instructive: in both cases, attackers targeted the human and procedural layer โ a signing interface in one, a withdrawal flow in the other โ rather than cryptography itself.
Here is how the Bitget attack unfolded, reconstructed from Chainalysis attribution and Bitget's own tracker. An attacker exploited an undisclosed flaw in a third-party security product. From there, they gained unusual familiarity with Bitget's internal systems โ enough to manipulate the withdrawal process directly, rather than crack any cryptographic layer. The stolen funds then moved through THORChain, a permissionless cross-chain DEX, and touched NEAR Intents. Certora's independent analysis estimated that roughly $238 million had already flowed out even after withdrawals were frozen.
Chainalysis attributed the operation to North Korean actors. That single fact reframes everything downstream: every wallet, protocol, and OTC desk that touches these funds now sits inside the OFAC sanctions perimeter.
It is worth naming the trust model plainly, because the industry often obscures it. Bitget users do not hold keys. They hold promises โ that the exchange will custody assets safely, monitor flows accurately, and honor withdrawals on demand. Every one of those promises broke at once. There is no on-chain trust minimization to fall back on, no smart contract to audit, no code to fork. When the operator's internal process is manipulated, users have exactly one recourse: faith in the operator's honesty afterward. This was never a protocol exploit. It was an operational security failure โ a manipulation of human process, not code.
The most important technical detail is the entry point itself. Chen described the intrusion as originating from "a previously unknown vulnerability in a third-party product," and confirmed that the vendor has not been publicly identified. Think about what that means. A security product โ a tool that exists precisely to reduce risk โ became the attack surface. And the industry cannot audit the risk because it does not know whose product to inspect.
When a security vendor becomes the vulnerability, every exchange that shares that vendor inherits an invisible liability. This is the supply-chain problem that CeFi has spent a decade pretending not to have.
Equally troubling is the attacker's depth of system knowledge. Chen admitted she "cannot explain how the attacker gained knowledge of Bitget's systems," adding that she wished she could ask them directly. Preliminary investigations found no internal participation. So how does an external actor learn the precise timing and approval logic of an internal withdrawal flow? The honest answer points toward either a compromised supplier or a long, patient reconnaissance campaign. Neither is comforting.
Chen also conceded a defensive failure that deserves more attention than it received. "Alerts need to be interpreted correctly, and monitoring should respond faster," she said. The system fired warnings. The humans did not act on them in time. That is the defining signature of operational security failure โ not a broken lock, but a lock nobody checked.
The remediation steps are directionally sound: affected systems isolated, internal credentials reset, access to sensitive systems tightened, and additional approvals introduced for withdrawals. But none of it has been tested under fire. Certora's $238 million outflow estimate โ measured after withdrawals were supposedly frozen โ suggests that existing controls may still contain bypass paths.
Then there is the solvency question, and this is where I switch hats from analyst to the person who once had to answer these questions live. Bitget points to a protection fund exceeding $300 million, replenished after BTC was drawn down. Yet Chen acknowledged the theft exceeded that threshold. A fund smaller than a single realized loss cannot credibly cover the next, larger one โ and that is before we discuss the "more than $1 billion" in capital she referenced without an audit, a breakdown, or any indication of how much is liquid.
I have written before about the ethical pulse of the decentralized economy, and it beats fastest when transparency is absent. Here, the transparency gap is measurable. Bitget's own tracker shows roughly $1.07 million frozen โ 0.28% of the loss. Recovery is functionally zero, and the final bill will land on Bitget's balance sheet.
Chen's interview is itself a case study in what I call selective transparency โ a pattern I first encountered during the 2017 ICO era, when foundations would trumpet partnership announcements and bury token unlock schedules. She volunteers that profitability impact is "limited," that clients are returning, that capital exceeds a billion. She withholds the vendor's identity, the liquid portion of that capital, the actual deposit recovery, and the true recovery rate. The asymmetry is the message: what is disclosed is what flatters, and what is withheld is what would let us price the risk.
Community Pulse: The sentiment data tells its own story. When Bitcoin withdrawals reopened, several institutional clients pulled large sums โ a far more damaging signal than retail panic, because institutions migrate slowly and rarely come back. Chen insists "many clients have already returned," and notes ETH inflows exceeded outflows an hour after withdrawals resumed. But no deposit data, no retention figures, and no capital-return numbers were provided. Optimism without measurement is marketing, not evidence.
Ethical Impact: On decentralization integrity and community welfare, Bitget scores poorly โ not because it was attacked, but because it responded with selective transparency, sharing the flattering numbers and withholding the load-bearing ones.
Here is the angle almost nobody is pursuing. The most dangerous entity in this entire episode is not the attacker. It is the unnamed vendor.
If that third-party product is used by multiple exchanges โ a near-certainty, given how concentrated the security-tooling market has become โ then a single unpatched flaw represents a systemic, industry-wide exposure that no one can assess until the vendor is named. Building bridges in a fragmented digital frontier means admitting that our supply chains are shared, and that one cracked support beam weakens every span.
The reason for the silence is easy to guess, and it is not purely investigative. Naming the vendor invites legal liability questions and could trigger a cascade of withdrawals at every exchange that depends on the same tool. So the industry sits in a prisoner's dilemma: everyone benefits from shared threat intelligence, and everyone fears that disclosing their own weaknesses will be weaponized against them. Chen hinted at exactly this tension when she proposed an "exchange-plus-protocol alliance," while admitting it is hard to share "enough information, but not too much."
The second blind spot is the permissionless protocol layer. Chen declined to blame THORChain, but asked its operators to explain what intervention might be possible, and pointed to NEAR Intents as an example of infrastructure that acts "when allowed." That is a quiet acknowledgment that the line between censorship resistance and money laundering is now a live regulatory battleground โ one where OFAC sanctions and North Korean attribution leave permissionless protocols with very little room to maneuver.
The next test for Bitget is not recovery โ that ship has sailed. It is whether the new withdrawal controls can stop a forged internal transfer, because a second breach would be existential. The most useful thing the exchange could publish tomorrow is a real proof of reserves: cryptographic, third-party verified, showing what is actually liquid. Bitget has six months to prove that its remediation was real and not rhetorical โ and the entire CeFi sector is watching, because the same unnamed vendor may be sitting inside their own stack. Until that changes, the $387.5 million figure is not the number to watch. The number to watch is how many other exchanges quietly depend on the vendor that still has no name.