By our count, MetaMask Staking queued 7,066 Lido validators to exit hours before its notice, naming no cause

Every running validator was queued to exit before MetaMask said anything

Every one of the 7,066 running validators among the keys that Lido's registry lists under the operator named Consensys had an exit message in an Ethereum block hours before MetaMask's first time-stamped public word about its incident. The first message was included at 11:53:59 UTC on 30 September (12:53 BST). The last was included at 17:28:59 UTC (18:28 BST). MetaMask's notice page carries a date but no time, so the earliest time-stamped notice we can find is its post on X, at 23:38:56 UTC (00:38 BST on 1 October): 6 hours 10 minutes after the last exit message and 11 hours 45 minutes after the first. At 32 ETH each, the 7,066 validators hold 226,112 ETH, about 2.3 per cent of the ETH Lido reports as pooled. MetaMask's notice gives no number, no amount and no list.

What this does not establish matters as much. It does not say why the exits were ordered, whether any validator key was exposed, or whether anyone has lost money. MetaMask says it has "identified no immediate threat to MetaMask wallets"; Lido says the cost is "foregone rewards as well as possible downtime penalties". The chain cannot say who submitted the exits, only that each was signed with the validator's own key. The match between the registry's operator number 21 and MetaMask Staking is our inference, from the operator's name, from Lido's wording and from a 2023 exit that the chain matches exactly to this operator (see "The numbers the notices leave out"); neither MetaMask nor Lido gives the operator number. And the figure covers only validators run under Lido. MetaMask also sells pooled and validator staking outside Lido, and none of that is visible here.

What happened, in the order the clocks say

The chain and the notices give two different starts to this story, and the chain is earlier. Lido's forum post, created at 23:40:49 UTC on 30 September, says relevant validators "have begun the exit process". By then the whole set had been queued for more than six hours. The first figure draws the two records to scale.

Public and on-chain events in UTC, with UK time. Sources: beacon node API, X, research.lido.fi, metamask.io page metadata, Internet Archive, news timestamps. Read 10:35 to 11:00 UTC on 1 October.

Event	UTC (BST)	Source
First of 7,066 exit messages included in a block (slot 15,328,768)	30 Sep 11:53:59 (12:53)	Beacon chain
End of the first burst: 5,229 validators in 1 h 42 min	30 Sep 13:36:47 (14:36)	Beacon chain, derived
Second burst, 1,837 validators in 26 min; last message at slot 15,330,443	30 Sep 17:02:47 to 17:28:59 (18:02 to 18:28)	Beacon chain
MetaMask's post on X: security update, no immediate threat to wallets	30 Sep 23:38:56 (1 Oct 00:38)	X, time from the post ID
Lido forum post: precautionary out of order exits	30 Sep 23:40:49 (1 Oct 00:40)	research.lido.fi
Lido's post on X, same text	30 Sep 23:44:25 (1 Oct 00:44)	X, time from the post ID
The Hacker News publishes	1 Oct 05:10 (06:10)	RSS feed, 10:40 IST
MetaMask notice page: modified-time metadata; text identical in a 06:01 UTC archive capture and at 10:35 UTC	1 Oct 05:21:28 (06:21)	Page metadata, Internet Archive
BleepingComputer publishes	1 Oct 07:33 (08:33)	Page stamp, 03:33 EDT

Two features of the chain record are worth drawing out. The messages came in two bursts, with 3 hours 26 minutes between them in which 2,432 exit messages for other validators were included, so the exit queue is shared with other stakers. And nothing has been added since: from 17:29 UTC on 30 September to our last scan at 11:09 UTC on 1 October, no further exit message for this operator appears in any block. The on-chain action was complete before the first public word.

MetaMask's page is dated 30 September and gives no time. In the UK its first post and both of Lido's posts fall just after midnight on 1 October, so a UK reader will find no time-stamped statement from either organisation before 00:38 BST on the 1st. From 23:38:56 UTC to 11:10 UTC (12:10 BST), MetaMask has published no second post on X, the notice text is unchanged from the 06:01 UTC archive capture, and Lido's forum thread has one post and no replies. The notice promises "further updates as appropriate".

What the two statements say, and where they differ

MetaMask's notice is six sentences. It says the company is "responding to an ongoing security incident affecting part of our infrastructure", is remediating it "in coordination with external partners and security advisors", has "identified no immediate threat to MetaMask wallets", is "proactively exiting affected validators" as a precaution, and that its staking is non-custodial because "we do not manage withdrawal keys for stake on behalf of our clients". It gives no number, no system, no start time and no instruction to anyone.

Lido's post, written after what it calls "an investigation into an infrastructure compromise", adds what MetaMask does not: the word compromise, a cost ("foregone rewards as well as possible downtime penalties"), a timetable ("by the end of October 7th, 2026"), a return of the ETH over "approximately up to 45 days due to the extended entry queue", and a reserve fund of "over 6,750 stETH". It tells stETH holders "No action is required". It refers readers to MetaMask for details, and MetaMask has given none.

The two organisations describe the same event differently: MetaMask says an incident affecting infrastructure, Lido says a compromise. Neither says what was compromised. Asked by BleepingComputer which part of its infrastructure was affected and whether any systems or data were accessed, a MetaMask spokesperson pointed the reporter back to the public statement.

One word is shared and undefined: "affected". MetaMask is exiting "affected validators". The chain shows the exits covered every one of the 7,066 still running in Lido. So either every validator was treated as affected, or the exit was an operator-wide precaution. The statements do not say which, and it matters: the first would put the whole fleet inside the incident.

The numbers the notices leave out

Ethereum is a public ledger, so some of what a notice omits can be computed. The operator numbers come from Lido's NodeOperatorsRegistry contract, read through a public RPC at block 26,096,908 (10:35:47 UTC). Its signing keys, 11,213 in all, were matched against the beacon chain's validator set, and the 8,594 slots from 06:00 UTC on 30 September to 10:38 UTC on 1 October were scanned for voluntary exits. Everything marked derived is our arithmetic.

What the chain shows about the operator Lido registers as Consensys (operator 21). Sources: Lido registry and stETH contract via public RPC; beacon node API (second node used as a cross-check); Coinbase spot price. Read 10:35 to 11:00 UTC on 1 October, status rows re-read at 11:10 UTC (12:10 BST).

Figure	Value	Where it comes from
Keys deposited under operator 21	11,213, of which 4,009 had exited before 30 September	Lido registry
Validators on the beacon chain	11,075; the other 138 keys (4,416 ETH) are still in the deposit queue	Beacon node; 11,075 plus 138 equals 11,213
Queued to exit on 30 September	7,066 (5,229 plus 1,837), which is every validator not already exited	8,594 slots scanned; derived
ETH held by the 7,066	226,112 ETH, about $610 million or £462 million	Derived: 7,066 x 32 at $2,696.84 and £2,041.41 per ETH, Coinbase spot, 10:47 UTC
Share of ETH Lido reports as pooled	2.3 per cent of 9.84 million ETH	stETH contract, 11:00 UTC; derived
Where the withdrawals are paid	All 11,075 carry Lido's withdrawal address, not MetaMask's	Beacon node; matches the stETH contract's withdrawal credentials
Slashed validators	0 of 11,075	Beacon node
Share of the whole exit queue	6,754 of 23,016 exiting validators on the network (29.3 per cent) at 10:47 UTC	Beacon node; derived
Status now	4,009 withdrawn earlier; 344 exited and waiting; 6,722 still active and exiting; none active and not exiting	Beacon node, 11:10 UTC (12:10 BST)
Lido's reserve against the exposure	Over 6,750 stETH, about 3.0 per cent of 226,112 ETH	Lido forum post; ratio derived

How long the exit takes, computed from the protocol

Ethereum limits how fast validators can leave. The beacon node reports an exit limit of 256 ETH per epoch, which is eight 32 ETH validators every 6.4 minutes, 1,800 a day. Alone in the queue, 226,112 ETH would need 883.25 epochs, 3 days 22 hours. It was not alone. The last of the 7,066 is scheduled to exit at epoch 480,501, which is 01:26:47 UTC on 7 October (02:26 BST), 6 days 13 hours 33 minutes after the first message. That sits inside Lido's "by the end of October 7th". Our count by day is in the second figure, and it matches the bursts exactly: the first burst's 5,229 validators fill 1 to 4 October (1,241, 1,228, 1,596 and 1,164) and the second burst's 1,837 take 6 and 7 October (1,733 and 104). 5 October has none of this operator's exits at all: the queue is first come, first served, and 2,432 other validators' exit messages were included between the two bursts, which is consistent with that day's capacity going to them (our inference).

The exit is not the end. A validator can be withdrawn only 256 epochs, 27.3 hours, after its exit epoch, then waits for the withdrawal sweep, which we did not compute, and the ETH returns through the deposit queue. That queue held 20,935 pending deposits totalling 1,602,437 ETH at 10:55 UTC, and the same at 11:10 UTC (12:10 BST). At 256 ETH per epoch a deposit joining now waits 27.8 days. At that length, re-entry of the first exited stake would start around 30 October and the last around 5 November, before the sweep and Lido's own deposit timing. Lido's "approximately up to 45 days", about 14 November if counted from its post, is a ceiling and does not conflict with that. The queue's own history agrees that it can be longer: the oldest deposit still waiting was submitted 37.9 days ago, on 24 August.

Lido's deposit queue also holds 138 more keys of this operator, 4,416 ETH, submitted between 25 August and 17 September. They sit at positions 461 to 14,909 of the queue, which at today's pace means activation in about 1.2 to 18.5 days. Neither notice says what is planned for validators that would start running on the same operator's infrastructure after the exit.

Two friendly labels: infrastructure incident and non-custodial

Both labels are accurate and both are asked to carry more than they can. A comforting phrase is not a control, and neither of these tells a staker or a security lead what to check.

Phrases in the two statements and on MetaMask's product pages, and what each does and does not establish. Sources: metamask.io notice and earn page, Lido forum post, MetaMask Staking disclosures (last updated December 2023), June 2024 pooled staking post. Read 1 October.

Phrase	What it establishes	What it does not
A security incident affecting part of our infrastructure (MetaMask)	A security event, in some systems, still under way when posted.	Which systems, how it started, when, whether data or keys were touched, or whether those systems are shared with the wallet or Infura.
An infrastructure compromise (Lido)	Lido's word, after an investigation, for the event.	Any detail. Lido points back to MetaMask's release, which has none.
Non-custodial; we do not manage withdrawal keys	The operator cannot move the principal: all 11,075 validators pay out to Lido's address.	Anything about signing keys, which the operator does hold. Those keys queued 226,112 ETH of stake for exit in under six hours and, if misused, could cause slashing or downtime.
No immediate threat to MetaMask wallets	No threat to the wallet product had been identified at the time of posting.	That the wallet is unaffected for good: the sentence is limited by immediate and by at this time, and the wallet is a different product from staking under the same brand.
Zero slashes since 2020; ISO 27001 and SOC 2 Type II (product pages)	A track record and audited management systems. The chain shows no slashed validator in this Lido set.	That any of it protects against this event. Exits are not slashing, and a clean slashing record is silent on whether keys were exposed this week.

Take non-custodial first. It is true on the chain: every one of the 11,075 validators carries Lido's withdrawal address, so MetaMask cannot redirect the principal. What it does not cover is everything else a signing key does. MetaMask's own disclosures page, last updated in December 2023, says a user "alone" accepts responsibility for losses "regardless of how they occur". The product page says "Unstake at any time". The protocol decides when: for this set the exit queue alone runs to 7 October, and the full cycle may pass a month. "Any time" means any time you ask, not any time you are paid.

What is stated and what is not, at 11:10 UTC on 1 October

Questions a defender would ask, with what MetaMask, Lido and the chain state and what they do not. Sources: MetaMask notice and X post, Lido forum and X posts, BleepingComputer, beacon chain. First read about 10:40 UTC and re-read at 11:10 UTC (12:10 BST).

Question	Stated	Not stated
What is affected	Part of MetaMask's infrastructure; its non-custodial staking operations; validators it operates for Lido. No immediate threat to MetaMask wallets.	Which systems. Whether Infura or any other service is involved. Infura's status page read all systems operational at 11:06 UTC, which is availability, not security.
How many validators, how much ETH	Nothing in either statement.	Whether affected means all. The chain shows 7,066 and 226,112 ETH in Lido (derived), and nothing outside Lido.
What users should do	Lido: no action is required from stETH holders.	Any instruction for wallet users, pooled stakers or validator-staking clients. MetaMask: further updates as appropriate.
Funds lost or at risk	MetaMask: no immediate threat to wallets. Lido: foregone rewards and possible downtime penalties. No source reports a loss.	Whether any user funds are at risk. The size of the penalties.
Cause and start	Lido: a compromise, after an investigation. MetaMask: a security incident, ongoing when posted.	How access was gained, when it began, when it was found, whether data was accessed.
Whether keys were exposed	MetaMask does not manage withdrawal keys; the chain confirms Lido's address on every validator.	Anything about signing keys. The chain shows the exits were signed by them.
Who did it	No actor is named by MetaMask, Lido, The Hacker News or BleepingComputer.	Any actor, or evidence for one.
Whether staking and the rest are separate	MetaMask's notice separates the wallet from the staking action.	Whether staking shares systems, credentials or staff with the wallet or Infura, or which company runs it after the 9 September split of Consensys Software Inc.
When the cause will be known	Lido: a full investigation is underway. MetaMask: updates as appropriate.	Any date. In Kiln's 2025 case the postmortem came 105 days after detection.

Read the middle column as the whole of the public record. The two rows a customer needs first, how many validators and what to do, are the two where the companies say least, and the only numbers in the table come from the chain, not from either company.

Two earlier exits in the same ecosystem, and the other incident this week

Earlier out of order exits recorded on Lido's forum. Sources: research.lido.fi topics 5278 and 10654, and the Lido registry. A difference in cause is stated where the record states one.

Event	What the record says	Cost and timing
Consensys Staking, 17 July 2023	125 validators exited by mistake; added safeguards.	About 39 days out of the active set; 19.732 ETH offered.
Kiln, September 2025	A compromised engineer's access token, then cloud credentials and an altered API response; treated as a full infrastructure compromise; exit began 9 September.	Lido analytics: 207.312 ETH of missed rewards on 5,726 validators, about 13.5 days each. Postmortem posted 22 December, 105 days after detection on 8 September.

The Kiln record is the closest precedent for the shape of this week: a precautionary exit, a short notice, a cause that arrived months later. Whether MetaMask's cause resembles Kiln's is unknown and should not be assumed. The registry shows Kiln has since exited all 10,579 of its Lido validators, so the precedent also shows how an operator leaves.

The other crypto compromise we covered this week is Bitget's: see the forensic reports of 30 September and the root cause described three days late. Bitget is an exchange and this is a staking operator. No source we read links the two, and we do not. What they share is a wording habit, the name of a place in place of a cause. Our earlier piece on Bitget's claim that its private keys were never compromised is about the same gap between a statement about keys and what a key did.

What to do, in the order worth doing

The question this leaves

MetaMask says it does not manage withdrawal keys. The chain shows what it does manage: signing keys, which put 226,112 ETH of other people's stake into the exit queue in under six hours. If an operator can do that before it speaks, the question for any staking provider is not whether it is non-custodial. It is which key the incident was about, and who will tell you before the chain does.

Key facts