On September 24, 2026, Bitget detected unauthorized cryptocurrency transfers from part of its hot and warm wallet infrastructure. By early October, the exchange said the affected funds totaled about $388 million.
The incident was large enough to be described simply as a crypto exchange hack, but that shorthand hides several different questions. Was the blockchain itself compromised? Were private keys stolen? How were withdrawal controls bypassed? And once assets left the wallets, who ultimately bore the financial loss?
Bitget’s latest explanation says the attacker exploited a zero-day vulnerability in a third-party security product, obtained high-privilege credentials and injected forged withdrawal instructions that bypassed risk controls. The company says private keys were not compromised and cold wallets were unaffected.
Updates from SlowMist and Mandiant broadly support a path in which the attack crossed a trust boundary between an external security environment and Bitget’s production wallet infrastructure. The investigation was still described as ongoing, and some vendor details remained undisclosed, so the causal account should be read as the best available picture as of October 6, 2026 rather than the final word.
For users, the more consequential question is what happened after the technical failure. Bitget said customer balances were not reduced and that the financial impact was absorbed through its Protection Fund. Withdrawals were restored in phases, with the remaining supported assets, fiat withdrawals and P2P scheduled to resume by October 2.
That distinction matters: the system that gets breached and the party that ultimately absorbs the loss are not necessarily the same thing.
Key insight: Asking whether a blockchain is secure is not enough when assessing a crypto exchange. Investors also need to ask who controls the keys, who can authorize transactions, which systems are trusted to issue instructions, and whether the platform has enough capital or contractual protection to absorb a loss if assets disappear.
“Crypto was hacked” can describe five very different risks
The phrase “crypto hack” often collapses several different failure modes into one label. Separating them makes the security question much clearer.
- Blockchain or asset protocol risk – vulnerabilities in a protocol, smart contract or bridge can allow network rules to be exploited in ways designers did not intend.
- Custody risk – weaknesses in the systems that hold private keys or control wallets on behalf of users.
- Operational security risk – failures around signing interfaces, privileged credentials, withdrawal workflows and third-party software connected to core systems.
- Counterparty risk – even after an attack has stopped, users remain exposed if the platform’s remaining assets are insufficient to cover its liabilities, creating liquidity, restructuring or insolvency risk.
- Regulatory and consumer-protection risk – customer rights differ across jurisdictions. Some legal regimes impose clearer asset-segregation and custodian obligations than others.
Key insight: A transaction can be fully valid from the network’s perspective while still being unauthorized from the user’s or institution’s perspective. The blockchain may simply execute a correctly signed instruction even when the process that created or approved that signature has already been taken over.
Comparing three crypto exchange breaches
| Case | Technical failure mode | Reported value affected | What happened to users |
|---|---|---|---|
| Bitget (2026) | A third-party security product was compromised, enabling privileged access and forged withdrawal instructions | About $388 million | User account balances were not reduced; Bitget said the impact was covered through its Protection Fund |
| Bybit (2025) | A multisig cold-wallet signing workflow was manipulated through the Safe environment in a supply-chain compromise | About $1.46 billion | The exchange filled the asset gap and restored 1:1 backing for the assets covered by its report |
| WazirX (2024) | A multisig wallet compromise created a major collateral shortfall | More than $230 million | The company entered restructuring; eligible users received a first distribution of roughly 85% of Approved Claims under the scheme |
Two questions should always be separated: where did the system fail, and who absorbed the loss? The first is a security-architecture question. The second depends on capital, liquidity, contracts and law.
Bitget: the attacker did not break the blockchain, but gained control around the keys
Bitget is a useful example of the difference between possessing a private key and controlling the workflow around it. According to the exchange’s latest account, the attacker did not directly steal the private key of a cold wallet. Instead, the attacker compromised a third-party security product and obtained high-level credentials.
In practical terms, the attacker gained enough authority to impersonate a trusted operator and generate forged withdrawal instructions that internal systems treated as legitimate. The lesson is that protecting the key itself is not sufficient if surrounding systems with authority over the key can be compromised.
Modern custody architecture therefore has to look beyond the simple hot-wallet-versus-cold-wallet split. It must also manage signer permissions, separation of duties, transaction policies, withdrawal limits and the security of external vendors. An asset is only as secure as the weakest privileged path that can move it.
When those technical controls failed, Bitget used its balance sheet and Protection Fund as the next layer of defense. The company said user balances would not be written down. Not every exchange can necessarily do the same, but the case illustrates how an institution’s ability to absorb losses becomes an economic security control after a technical one breaks.
Key insight: The critical question is no longer only “is the key safe?” but also “who controls the systems that are allowed to act around the key?” Even when the blockchain and private key are not directly compromised, control of the surrounding workflow can still produce the same economic result.
Bybit: a cold wallet can still be vulnerable if the approval process is deceived
On February 21, 2025, Bybit lost assets from an Ethereum multisig cold wallet. The exchange put the loss at about $1.46 billion, while the FBI cited roughly $1.5 billion and attributed the theft to actors linked to North Korea.
The case shows why the label “cold wallet” does not eliminate risk. If the process that signers use to inspect and approve a transaction can be manipulated, the wallet can still authorize a transaction the human operators never intended.
Bybit said the incident occurred during a multisig transaction through Safe, where information presented to signers did not match the transaction ultimately executed on-chain. The control problem therefore extended beyond the number of signers to the software stack, the verification interface and the ability to independently confirm destinations and transaction data.
After the incident, Bybit said it closed the ETH deficit through short-term borrowing, OTC purchases and large deposits. A subsequent proof-of-reserves report said in-scope customer assets were again backed 1:1 within about 72 hours.
Key insight: Offline key storage does not guarantee safety when the signing workflow can be manipulated. And a proof-of-reserves snapshot confirming 1:1 backing after an incident is useful evidence, but it does not by itself prove long-term solvency, legal segregation of customer assets or future security.
WazirX: when a cyber loss becomes a liquidity and creditor problem
WazirX shows a different endpoint. After a multisig wallet attack on July 18, 2024, the company said more than $230 million in digital assets had been stolen and acknowledged that the loss affected its ability to maintain 1:1 collateral backing.
At that point the problem stopped being only an incident-response question. It became a balance-sheet question: how would platform liabilities be met when the remaining assets no longer matched what was owed to users?
In 2025, WazirX operator Zettai entered a Scheme of Arrangement in Singapore. The company said the scheme became effective in October 2025. Eligible users then received a first distribution representing about 85% of Approved Claims, calculated under the scheme’s reference-pricing methodology, together with Recovery Tokens that may participate in future recoveries.
The 85% figure therefore should not be read as “85% of the original coins returned at today’s market price.” It is a distribution of value under the legal restructuring framework.
Comparing Bitget, Bybit and WazirX makes the economic distinction clear. Incidents that look similar in headlines can produce very different outcomes for users.
If an exchange has enough capital, liquidity or reserve resources, users may avoid a direct haircut. If the loss punches through the balance sheet and leaves assets below liabilities, the problem rapidly becomes one of insolvency, creditor priority and property law.
Key insight: A cyber incident does not end when the technical hole is patched. Once a loss breaches the institution’s financial defenses, it becomes a debt-structure problem, and user recoveries depend heavily on the platform’s balance sheet and the applicable legal framework.
Why blockchain funds can be traceable without being automatically reversible
One of crypto’s central paradoxes is that many transactions are recorded on a public ledger that lets investigators follow the flow of funds, yet visibility does not create an undo button.
Once a transaction satisfies network rules, reversing its economic effect generally requires an external control point. A centralized exchange receiving the assets may freeze them. A stablecoin issuer may have the ability to freeze or disable specific tokens. Law-enforcement agencies may seize assets when they have legal authority and cooperation from service providers.
The Bybit case produced examples of partial freezing and seizure. The T3 Financial Crime Unit announced that it had frozen nearly $9 million linked to the incident in March 2025, while Chainalysis later described a Greek law-enforcement seizure of additional assets tied to the theft.
These cases show that traceability can be valuable for enforcement, but they do not mean that most stolen funds are always recoverable. Nor do all digital assets have a centralized emergency switch.
For base-layer assets such as Bitcoin or Ether, there is no central issuer that can retroactively freeze a transaction in the way some stablecoin issuers can. The faster an attacker disperses assets across chains, swaps, bridges or services that do not cooperate with investigators, the harder recovery becomes.
A transparent ledger therefore helps identify the trail, but it does not turn a blockchain into a credit-card-style chargeback system.
A blockchain can be transparent and final at the same time. Seeing where stolen assets moved does not automatically give anyone the authority to pull them back.
Self-custody reduces counterparty risk but transfers responsibility to the key holder
After an exchange hack, a common recommendation is to move assets into self-custody. That can reduce one class of risk because the user, rather than an exchange, holds the private key. Exposure to an exchange bankruptcy, withdrawal freeze or custody failure therefore falls.
But the risk does not disappear. It moves from an institution with security teams, incident-recovery procedures and a balance sheet to the individual user. The owner must protect the seed phrase, verify destinations, defend against phishing, separate devices where appropriate and plan for recovery if access is lost.
Chainalysis reported that personal-wallet compromises accounted for 23.35% of the stolen-fund activity it tracked at midyear 2025. That figure does not prove self-custody is more dangerous than exchange custody because the definitions and data sets are different. It does show that moving assets away from an intermediary changes the attack surface rather than eliminating it.
The better question is therefore which risks a user is prepared to manage personally and which capabilities they expect from a custodian.
For exchange users, security due diligence should go beyond claims about cold wallets or proof of reserves. It should include signer governance, segregation of customer assets, withdrawal controls, reconciliation and audit practices, insurance or protection funds, legal terms and the platform’s history of handling incidents.
Key insight: Self-custody is not a cure-all. It replaces platform risk with personal operational responsibility. If assets remain with an intermediary, investors need to look past technical marketing and examine governance, segregation and the institution’s capacity to absorb losses.
The future of digital money may be less about eliminating trust than knowing what we trust
The original promise of blockchain technology often begins with trust minimization: reduce reliance on intermediaries through verifiable rules and user-controlled keys. In the real economy, however, most users still depend on exchanges, custodians, stablecoin issuers, wallet software, cloud infrastructure, cybersecurity vendors and legal systems.
Trust has not disappeared. It has been redistributed across layers of the ecosystem.
Bitget, Bybit and WazirX show that digital-money security is a socio-technical problem rather than a cryptography problem alone. A strong protocol still depends on operational discipline, independent verification, asset segregation, incident response, liquid reserves and clear rules for allocating losses.
The reverse is also true: a strong intermediary cannot eliminate protocol risk or mistakes made by users themselves.
For users, the question should move from “has this exchange ever been hacked?” to “if one control layer fails, can the next layer contain the damage?” That means examining signer permissions, withdrawal limits, independent verification, segregation of customer property, the balance sheet’s capacity to absorb shocks and the customer’s legal rights in a worst-case scenario.
The lesson extends beyond cryptocurrency. More forms of money and financial assets are becoming dependent on software and custody infrastructure, from tokenized assets and stablecoins to new digital payment systems.
Credibility therefore does not come from choosing between “trust the blockchain” and “trust the institution.” It comes from designing layers that can verify one another and absorb failures without automatically pushing the entire cost onto end users.
Key takeaways
- An exchange hack does not necessarily mean the blockchain failed. Losses can originate in custody systems, signing processes, privileged access, withdrawal controls or third-party software while the blockchain continues to execute valid instructions.
- Where the breach occurred and who paid for it are different questions. Bitget and Bybit said they absorbed the deficit, while WazirX moved into restructuring after the loss created an asset shortfall.
- Traceability is not reversibility. Public ledgers can help follow stolen assets, but freezing or recovering them usually requires a centralized control point, cooperation from service providers or legal authority.
- Self-custody transfers risk rather than removing it. It reduces exchange counterparty risk but makes the user responsible for key management, phishing defense and recovery planning.
- The future of digital money needs both protocol security and institutional resilience. Secure keys must sit inside systems with sound governance, asset segregation, risk controls, liquidity and clear customer rights when a crisis occurs.
