1 BTC ≈ 64,519 USDT1 ETH ≈ 1,810 USDT1 XMR ≈ 486 USDTmedian across 6–8 services · tick 60s
home / blog / bitget-hack-387m-cross-chain
Blog

Bitget hack: how $387.5M moved through cross-chain swaps — and what it means if you swap without KYC

Bitget lost $387.5M on Sept 24, 2026. How attackers got in, how the funds moved through THORChain and NEAR Intents, and how to avoid tainted coins.

by Milan Vosspublished updated 12 min read

What happened at Bitget

Bitget is a large centralized exchange that requires KYC — it is not a service we review. We cover this breach because of where the stolen funds went next: through the same cross-chain swap infrastructure our readers use every day.

date (UTC)event
Aug 31The earliest malicious activity SlowMist and Mandiant found in Bitget’s logs: a zero-day vulnerability in a service on a third-party security product (Cointelegraph).
Sept 24, 18:31–21:23Unauthorized transfers run for about three hours — 02:31 to 05:23 on Sept 25 in UTC+8, the time zone of SlowMist’s report. Bitget halts withdrawals and reports $351.6 million affected.
Sept 25Bitget revises the total to about $387.5 million after counting Zcash and TRON transfers, launches a recovery bounty and a live tracking dashboard. ZachXBT says he has no plans to monitor the case.
Sept 26–28THORChain declines Bitget’s request to refuse the attacker’s addresses; NEAR Intents reports blocking over $50 million in attempted transfers.
Sept 28ZachXBT publishes five accounts he says are laundering the funds. Bitget restarts Bitcoin withdrawals.
Sept 30Bitget restores BTC, ETH and USDT withdrawals and publishes a proof of reserves showing a 131% reserve ratio across 19 assets. It also says its protection fund is back above $300 million, restored with its own capital.
Oct 1Chainalysis attributes the theft to North Korea-linked actors and reports that $387 million left Bitget in 23 transfers during the first three hours.
Oct 2Bitget’s CEO tells CNBC that about $1.1 million has been frozen and that she does not expect to recover much. The last withdrawal phase — other tokens, fiat and P2P — is scheduled for 08:00 UTC.

The affected assets included ETH, XRP, BNB, AVAX, USDT and USDC on Ethereum, XRP Ledger, Arbitrum, Avalanche, Optimism, BNB Chain and Base, CEO Gracy Chen said.

How the attackers got in

This was not a stolen-key hack. According to Bitget’s own account, relayed by Fortune and BleepingComputer, and the findings of the two firms it hired, SlowMist and Mandiant:

  1. The attackers exploited a zero-day vulnerability in third-party security software used by the exchange. The earliest malicious activity in the logs dates to August 31 — more than three weeks before any money moved.
  2. They reached the management platform of a second security product by impersonating an internal employee, and obtained internal credentials.
  3. Mandiant found that on September 24 the attacker had privileged access to both security appliances and, through one of them, moved to the production wallet job server and deployed malicious packages.
  4. With a custom withdrawal tool they forged transfer data so that fraudulent withdrawals looked legitimate, and Bitget’s approval process signed them. SlowMist also found attempts to tamper with withdrawal records and to trigger additional BTC withdrawals (Odaily).
  5. The transfers came only from hot and warm wallets. Bitget says its cold wallets and the separate Bitget Wallet product were not affected.

Bitget says the flaw is fixed and that there have been no further unauthorized transfers (The Block).

The lesson is uncomfortable for every custodial service: an approval system is only as strong as the software it trusts. The keys were never touched — the process that decides when to use them was.

Who was behind it?

Bitget says it suspects North Korean attackers; its CEO cited IP addresses matching the VPN infrastructure of a North Korea-linked group. Since then, analytics firms have gone further: Elliptic assessed a link as highly likely, citing on-chain overlaps with earlier North Korea-attributed thefts, and on October 1 Chainalysis attributed the theft to North Korea-linked actors, saying it pushes thefts by North Korea-linked groups in 2026 past $1 billion. These are assessments by private firms. As far as we can find, no government has formally attributed the attack.

What has been published concerns the laundering, not the breach itself. On September 28, ZachXBT named five accounts he says are moving the stolen funds on behalf of the suspected attackers, and matched each to a transaction. He says they asked for help with stuck transactions in the public Discord and Telegram channels of services they were using. One of them — known as “lolo”, also referred to as “Alias 4” — also laundered funds from the $292 million Kelp DAO exploit in April. ZachXBT says he has seen the same pattern after several exploits attributed to the North Korea-linked group known as TraderTraitor, and plans to publish more data.

As one analysis notes, this evidence does not by itself prove who carried out the breach.

Where the money went

The stolen funds were split across chains and services. ZachXBT and others report movement through cross-chain bridges and swap protocols, including THORChain, and into the bitcoin mixer Wasabi.

Two kinds of freezes have happened so far:

  • Stablecoin issuers froze about $318,000: Tether roughly 218,000 USDT and Circle about 100,000 USDC.
  • NEAR Intents froze about $503,000 during execution and rejected the rest of more than $50 million it flagged.

That is less than 0.3% of the total.

Update, September 30: ZachXBT reports that about 18,900 ZEC (around $28.3 million) of the stolen funds came from Bitget’s Zcash hot wallet, and that the suspected attackers began moving about 2,700 ZEC into Zcash’s shielded Ironwood pool. Blockchain tracker MistTrack reported on October 1 that addresses linked to the attackers were still converting DAI and bridging it to TRON.

Update, October 2 — where the money went. Blockchain security firm BlockSec traced the funds using Bitget’s published attacker addresses and a September 29 snapshot (BlockSec, as reported by crypto.news):

  • First 41 minutes: about $75.5 million in freezable stablecoins (USDT, USDC, USDT0) was swapped into ETH and AVAX through Uniswap, UniswapX, 1inch and MetaMask’s built-in swap — before Bitget announced the breach.
  • Cross-chain: about $269 million in pass-through value moved through THORChain (7,804 transactions) and about $37.3 million through Chainflip; other routes included USDT0/LayerZero, Circle’s CCTP, Across and Stargate. The figures overlap: the same funds can pass through several protocols.
  • Still with the attacker: about $342 million (88.3% of the total) five days in, roughly 83.7% of it in bitcoin (about 3,386 BTC). About $3.9 million had entered CoinJoin transactions.
  • Frozen in public: about $840,000 — 0.2%. Tether and Circle immobilized about $340,000; NEAR Intents stopped about $503,000 during execution while about $166,000 passed through (its own disclosure gives an error margin of up to 10%).
  • THORChain declined Bitget’s request to reject the attacker’s addresses, saying its emergency halts protect the protocol and are not selective freezes.
  • BlockSec cautions that the overlaps with earlier DPRK-attributed attacks point to reused laundering groups rather than proof of who the attacker is; no law-enforcement agency has attributed the breach.

For readers comparing cross-chain services: Chainflip lets brokers and whitelisted screeners reject flagged deposits, and about $37 million still passed through it — screening narrows a route, it does not close it. Gracy Chen has said she is not very optimistic about recovery: once funds have crossed chains, freezing gets hard.

Update, October 8. Chainalysis says the first wave — $387 million in 23 transfers over three hours — landed on Ethereum (49.7%), the XRP Ledger (40.8%), Zcash (7.6%) and Tron (1.8%), and that tens of millions of dollars in stolen XRP were fed into cross-chain liquidity protocols and came out as bitcoin over about a day and a half (CoinsPress, summarizing the report). In the same coverage, TRM Labs traced BNB and TRX through THORChain and smaller amounts through Across, Bridgers, Chainflip and FixedFloat. FixedFloat is one of the services in our catalog; sitting on a laundering route is not a finding about a service’s own conduct, and we note it only so that readers can see the route. About $1.1 million has now been frozen — under 0.3% — according to Bitget’s CEO; frozen is not the same as returned.

NEAR Intents vs THORChain: two answers to the same question

The Bitget hack turned a long-running debate into a live test. What should a cross-chain protocol do when stolen money arrives?

NEAR IntentsTHORChain
What it didIts screening layer SHIELD flagged and rejected more than $50 million in attempted transfers; froze about $503,000 during execution; about $166,000 passed throughDeclined Bitget’s request to refuse service to the attacker’s addresses
Stated positionWill actively fight the laundering of hacked funds; frozen funds to be returned through a legal process; waived Bitget’s 5% + 5% bountyIt is “decentralized and permissionless like Bitcoin, Ethereum, and BNB Chain”
Where screening sitsOn integrated quote flows, before the swap; its documentation says coverage can vary by flowNot at protocol level; interfaces built on top may screen
SourcesCointelegraph, NEAR Intents docsCointelegraph Magazine

Note what happened to the money NEAR Intents rejected: it moved to other providers. Screening at one door does not stop the money; it changes which door it uses.

THORChain has been here before. It was used to swap a large share of the funds from the $1.46 billion Bybit hack in 2025. Lawyers now argue over whether “decentralized” is a defense, and over whether NEAR’s blocking shows that its system can be made to block in other situations too.

What our own data shows

We label screening for every service in our catalog: what is checked, when, and by whom, quoted from each service’s own terms or documentation. As of October 8, 2026:

  • Cross-chain protocols: Relay, NEAR Intents and Garden screen addresses before the swap starts; deBridge checks the country you connect from. THORChain and Maya do not screen at protocol level. At Chainflip, brokers — and whitelisted screeners acting through any broker — can reject flagged deposits.
  • Instant exchanges: of the 20 in our catalog, 11 risk-score incoming coins after your deposit arrives — all 8 that are listed and 3 of the 12 still under review. Two check the country you connect from, one screens addresses, and for five we found no screening mentioned in their terms. For one we could not read the terms at all.

In other words, “no KYC” rarely means “no checks”. The checks have moved: from your identity to your coins. See the catalog — each service shows its screening label, with the full quote and source on its review page.

If you swap without KYC: how to avoid tainted coins

After a hack of this size, stolen coins spread. Most of them will never touch you, but here is how to keep it that way.

Before you swap

  • Know where your coins came from. If someone sends you funds with no clear history, do not swap them straight away.
  • Read the screening label of the service. A service that screens before the swap rejects you early; one that screens after your deposit can hold your coins.
  • Always set a refund address when the service allows it, so a rejected swap comes back to you, not into a support queue.
  • Check large incoming payments against public lists. Bitget published an attacker-address dashboard and API; ZachXBT’s posts list transaction hashes.

If your swap is held

  1. Do not send more funds to “unlock” it.
  2. Ask support in writing which clause of their terms they rely on.
  3. Keep your order ID, deposit and payout transaction hashes, and screenshots of the quote.
  4. Read what the service’s own terms say about holds and refunds — we quote them, with paragraph numbers, in our frozen-funds guide.
  5. If you do not want to publish your evidence, you can still prove later that you had it today: submit only its sha256 hash in the comment form of the service’s review — we timestamp it in Bitcoin.

Will Bitget users get their money back?

Bitget says user balances were unaffected and that its user protection fund covers the loss. The fund held about $465 million when the incident was disclosed. It fell below $200 million, according to Bloomberg’s calculation from the fund’s published addresses, and Bitget says it has since restored it to more than $300 million (about $309 million) with its own capital. The fund is separate from the reserves behind customer balances, which Bitget reports at 131%; users can check their balances through its Merkle-tree tool. Withdrawals were restored in stages from September 28. The last phase — other tokens, fiat and P2P — was scheduled for October 2, and we could not find Bitget’s own confirmation that it completed, so check its status page.

Recovering the stolen funds themselves is another matter. Bitget’s recovery bounty pays 5% for helping freeze funds and another 5% for recovering them. About $1.1 million has been frozen so far, and Bitget’s CEO told CNBC on October 2 that she does not expect to recover much.

Frequently asked questions

How much was stolen in the Bitget hack? About $387.5 million. Bitget first reported $351.6 million on September 24, 2026, and raised the figure after counting Zcash and TRON transfers it had missed; it says the increase reflects fuller accounting, not a second theft.

Were Bitget users’ funds lost? Bitget says user balances were unaffected and its protection fund covers the loss. The fund held about $465 million when the incident was disclosed, fell below $200 million according to Bloomberg’s calculation, and has been restored to more than $300 million with Bitget’s own capital. Withdrawals were restored in stages from September 28; the last phase was scheduled for October 2.

Was North Korea behind the Bitget hack? Bitget suspects North Korean attackers, and analytics firms now agree: Elliptic calls a link highly likely, and Chainalysis attributed the theft to North Korea-linked actors in a report on October 1. As far as we can find, no government has issued a formal attribution. ZachXBT’s published evidence concerns the people laundering the funds, not the people who carried out the breach.

Has any of the stolen money been recovered? Very little. Bitget’s CEO said on October 2 that about $1.1 million had been frozen — under 0.3% of the total — and that frozen funds had not necessarily been returned to the exchange. She said she does not expect to recover much.

Why did NEAR Intents block the funds while THORChain did not? NEAR Intents runs a screening layer called SHIELD on its quote flows and rejected more than $50 million in attempted transfers. THORChain says it is decentralized and permissionless and does not selectively refuse transactions.

Could I receive stolen Bitget coins in a no-KYC swap? It is possible if you swap with someone who received them. Most instant exchanges we track risk-score incoming coins after your deposit, so tainted funds can lead to a hold or a request for source-of-funds information.

What should I do if my swap is held after a big hack? Do not send more funds to unlock it. Ask support in writing which clause they rely on, keep your order ID and transaction hashes, and read what that service’s own terms say about holds and refunds.

Sources

This article is updated as the investigation continues; last updated October 8, 2026. We do not review Bitget and have no commercial relationship with it, THORChain or NEAR Intents.