Advertisement

Could AI answer your customers from your own documents?

Assistants trained on your files, website and FAQs.

Hack Database

MetaMask: Hackers Took Under $1,000, So It Pulled $1.4B of Ether Offline

An intruder in MetaMask's staking business redirected 0.36 ETH of tips. MetaMask answered by retiring about 17,000 validators holding roughly $1.4B.

MetaMask: Hackers Took Under $1,000, So It Pulled $1.4B of Ether OfflineLogo: MetaMask
MetaMask says no customer wallets or money were touched.

Key numbers

  • ~$970 (0.36 ETH)

    Stolen

  • ~17,000

    Validators exited

  • 523,000+ (~$1.4B)

    ETH pulled offline

  • Unknown (Tornado Cash-funded wallet)

    Attacker

At 10:27 UTC on 30 September 2026, a new Ethereum wallet received a little under 0.1 ether from Tornado Cash, a service that mixes coins so their origin cannot be traced. Nineteen minutes later, MetaMask's staking business began switching off the machines it runs on Ethereum for other people. By the next morning it had pulled about 17,000 of them, holding more than half a million ether, out of service.

The thief's haul from all this was 0.36 ether, about $970. The money MetaMask took offline as a precaution was around 523,000 ether, roughly $1.4 billion, by the count of the security researcher 0xKaden. Bitquery, a blockchain data company, later put the figure at 16,965 validators and 565,056 ether, about $1.5 billion, and called that a floor. No customer's money was taken. The cost falls instead on everyone whose ether was earning interest through those machines and now earns nothing for several weeks.

MetaMask is the browser and phone wallet that tens of millions of people use to hold crypto. Its parent company was called Consensys until 9 September, when it announced a split: the wallet and consumer business would take the MetaMask name under the co-founder Joe Lubin as chief executive, while the protocol work, including the Linea network, would move to a new company keeping the Consensys name under Mike Kriak. The staking arm, formerly Consensys Staking, now trades as MetaMask Staking.

Advertisement

Free AI assistant

AI that works, and proves it.

Ask Cynt how AI could help your business. Free to try.

Free AI assistant

Clerks, stamps and tips

Ethereum has no central bank or clearing house. Its transactions are checked and recorded by validators, which can be thought of as clerks. Each clerk posts a deposit of 32 ether, about $86,000, as a bond for good behaviour. Clerks take turns assembling the next batch of transactions, called a block, and the clerk whose turn it is collects the tips that senders attach to their payments. Running thousands of clerks is a business, and MetaMask Staking is one of the companies that does it for clients, earning them interest on their ether in the way a custodian earns a fee for lending out shares.

Each clerk has two keys. The signing key is the clerk's stamp: it approves batches and keeps the clerk working, so it has to live on an internet-connected computer. The withdrawal key is the key to the vault where the 32-ether deposit sits. Under MetaMask's arrangement, which it calls non-custodial, the clients hold the vault keys and MetaMask holds only the stamps. Both MetaMask and Lido stressed that point in their statements.

There is also a third, humbler setting: the fee recipient, the account the clerk's tips are paid into. That is what the intruder changed.

What the intruder did

Between 12:12 and 16:46 UTC on 30 September, 19 MetaMask-run validators built a block, according to 0xKaden's on-chain analysis, which neither company has confirmed. Eighteen of them paid their tips not to the client but to the Tornado Cash-funded wallet. Bitquery's reconstruction found the same 18 blocks and put the total at 0.36 ether. The biggest victim was Lido, the largest staking service on Ethereum, which pools ether from many savers and hands it to professional operators such as MetaMask: 11 of the diverted blocks came from Lido validators, worth about 0.19 ether. A block built at 12:11 paid Lido correctly. The next one, 48 seconds later, paid the intruder.

In stock-market terms, someone got into the back office of a custodian and changed the bank details on the commission statements. The shares never moved. The 32-ether deposits were never at risk of withdrawal, because the withdrawal keys were never in MetaMask's systems to steal.

MetaMask did not wait for the diverted tips. Its first exit request went out at 10:46 UTC, about 85 minutes before the first diverted block, which suggests the company had spotted the intrusion before the thief had collected anything. It reset the tip accounts in waves between 16:10 and 17:08 UTC. The last diverted tip arrived at 16:46.

MetaMask has not said how the intruder got in, and the blockchain cannot show it. Bitquery notes that the earliest message naming the intruder's wallet was signed a few minutes before the first diverted block, which means either the signing keys themselves were used or the settings on the signing machines were changed. It could not tell which.

Why exit over $970

Bar chart of exited validators: Lido main set 7,066, smaller clients 4,264, rewards to a Coinbase-labelled address 4,041, MetaMask pooled staking 1,469, Lido small set 125CMZ chart. Data: Bitquery
Whose validators were pulled offline.

The danger was never the tips. It was the stamps. A clerk who stamps two conflicting batches is punished by the network in a process called slashing, which burns part of the deposit and throws the clerk out. Anyone holding a validator's signing key can cause that on purpose. Bitquery estimated that if all 17,000 validators had been slashed at once, about 22,000 ether, roughly $59 million, would have been destroyed. 0xKaden said the attacker "likely never had the ability" to withdraw the staked ether, but that validators could in principle have been slashed, depending on how the access was obtained.

A stamp that may have been copied cannot simply be cancelled and reissued. On Ethereum the only way to retire a signing key is to retire the validator: exit, wait for the deposit to come back, and start a new validator with a fresh key. So MetaMask exited all of them. No slashing has been reported.

The statements

MetaMask posted its first update late on 30 September. It said it was responding to "an ongoing security incident affecting part of our infrastructure", that it had "identified no immediate threat to MetaMask wallets", and that it was "proactively exiting affected validators within our non-custodial staking operations". An update on 1 October said there was "no indication that MetaMask wallets or customer funds have been affected" and warned users never to share their recovery phrase. Asked by BleepingComputer which systems had been reached, a spokesperson pointed to the public statement.

Lido's notice went up on its research forum at 23:40 UTC, posted by Will Shannon. It said MetaMask Staking had acted "following an investigation into an infrastructure compromise", that the final validators would be exited by the end of 7 October, and that the round trip of exit, withdrawal and re-entry could take "approximately up to 45 days". It told holders of stETH, the receipt token Lido issues for staked ether, much like a unit in a fund, that no action was required, and pointed to a reserve fund of more than 6,750 stETH as a cushion.

Others moved to reassure their own customers. Stani Kulechov, founder of the lending protocol Aave, said Aave was watching alongside Lido and had seen no impact on its markets. Guy Young, founder of Ethena, said its USDe digital dollar had no direct stETH exposure.

The queue

Ethereum lets only a limited number of validators leave or join each day, so a mass exit becomes a queue. The ether waiting to leave rose from about 200,000 to more than 700,000 within a day, and the wait stretched from about three and a half days to nearly two weeks, according to The Cryptonomist and CryptoSlate. Besides more than 7,000 validators for Lido, Bitquery counted about 4,000 whose tips go to an address labelled as Coinbase's, about 1,500 in MetaMask's own pooled staking product and some 4,300 for smaller clients. Every one of them earns nothing until its ether is back in a new validator.

It has happened before. In September 2025 Kiln, another large staking company, exited all its Ethereum validators after an infrastructure breach that came days after the Swiss app SwissBorg lost about $41 million in Solana through Kiln's systems. MetaMask lost far less than $41 million. Its clients may still lose more than $970 in missed interest.

The aftermath

As of 13 October 2026, MetaMask has not published a post-mortem. It has not said how the intruder got in, whether the validator signing keys themselves were copied, or how many validators it exited in total. The figures in circulation, about 17,000 validators and between 523,000 and 565,000 ether, come from outside researchers, not from the company. The Tornado Cash-funded wallet that received the 0.36 ether had made no outgoing transactions as of 2 October. No one has been identified or charged.

Lido set 7 October as the date for the last MetaMask-run validators in its protocol to exit, and estimated that the ether could take up to 45 days to come back into service because of the queue to rejoin. On 5 October Lido contributors proposed, for the next on-chain vote of Lido's token holders, that no new deposits be allocated to MetaMask Staking. No validator was reported slashed, and no client lost any staked ether. The cost is the interest that more than half a million ether does not earn while it sits in queues, plus any small penalties for validators taken offline before their exits completed.

The incident came less than three months after Drop Site News reported that Consensys had unknowingly hired a developer linked to North Korea to work on MetaMask code in March and April 2026. Nothing published so far connects the two.

What this teaches

  • On Ethereum a suspect signing key cannot be replaced in place. The only fix is to exit the validator, so a small breach of the signing side can force a very large and slow exit.
  • Keeping withdrawal keys with the client worked as designed: the intruder could redirect tips but never move the 32-ether deposits.
  • The fee recipient is a quiet setting with real money behind it. Checking that every block pays the expected account would flag a diverted tip within seconds.
  • Staking services that spread ether across many operators, as Lido does, limit the damage from one operator's failure, but their savers still lose interest while the queue clears.
Advertisement

Powered by Cyntri AI

The technology behind this newsroom. Built by Cyntri AI.

AI-assisted publishing, content pipelines and analytics. Crypto Media Zone is our working example.

Powered by Cyntri AI