Ethereum Slashing Risk: Causes, Penalties and Prevention

By Owen Thurm Published November 6, 2026 5 min read

Quick answer. Signing two blocks for one slot, or conflicting attestations, gets a validator slashed. Since Pectra the first penalty is 1/4,096 of effective balance; a correlation penalty can then take everything if a third of stake is slashed together.

Watch: Ethereum Slashing Risk: Causes, Penalties and Prevention 1:41

Video transcript

Ethereum slashing rarely comes from attacks. In published post-mortems, it traces to one validator key signing from two places at once.

Three offenses get a validator slashed: a double proposal, a double vote, or a surround vote. Going offline is not one. It only costs rewards.

Since the Electra fork in May 2025, the initial penalty is small, one four thousand ninety sixth of effective balance. The validator is forced out and its stake stays locked for about 36 days.

The penalty to model is correlation, landing about 18 days later. Slashed alone, it is near zero. If one percent of stake is slashed, about three percent. At a third, the whole balance.

Pectra also allows compounding validators up to 2,048 ETH. One key can carry 64 times the stake of a 32 ETH validator, so one mistake slashes more.

Real incidents follow one pattern. RockLogic lost 11 validators when migrated keys came back. Launchnodes lost 20 in a failover. An Ankr cluster on SSV lost 39 during maintenance.

Prevention starts with one active signer per key. Keep slashing protection on every signer, move the signing history with each key, and destroy the old instance first. Doppelganger detection is a last line.

Read the full article for the complete checklist, or talk to Guardian about pressure-testing the systems behind your validator keys.

Slashing is Ethereum's penalty for a validator that signs conflicting messages. The first penalty is now small.

The real risk is correlation. When many validators are slashed in the same window, each one loses far more, up to its whole balance.

The slashings in published post-mortems did not come from attacks. They trace to one key running in two places, during a migration, failover or maintenance.

What Slashing Is

Validators stake ETH and sign blocks and attestations. Slashing punishes the signatures that could help rewrite the chain.

It is not a downtime penalty. Ethereum.org's DVT guide states that "going offline is not a slashable offense"; an offline validator only misses rewards.

A slashed validator is forced to exit, and its stake stays locked for about 36 days.

Double Votes and Surround Votes

The consensus specs define three slashable offenses:

Double proposal: two different signed blocks for the same slot.
Double vote: two different attestations with the same target epoch.
Surround vote: an attestation whose source-to-target span surrounds another's.

Anyone who sees two conflicting signed messages can submit them as proof. Once a block includes it, the protocol slashes the validator.

A single honest client never signs both messages. It takes the same key signing from two places, or a client that has lost its signing history.

The Penalties Since Pectra

The Electra fork, the consensus half of Pectra, changed the numbers on May 7, 2025. The Fulu fork of December 3, 2025 left them unchanged.

Penalty	Before Electra	Since Electra (May 7, 2025)
Initial penalty	1/32 of effective balance: 1 ETH at 32 ETH	1/4,096: 0.0078125 ETH at 32 ETH
Correlation penalty	Effective balance × 3 × share of stake slashed in the surrounding 36 days	Same formula, finer rounding
When it lands	About 18 days after slashing	Same
Missed duties	Penalized as offline until withdrawable	Same
Forced exit	Withdrawable after at least 8,192 epochs, about 36 days	Same
Maximum loss	The full effective balance	Same

The correlation penalty is the one to model. It scales with how much of all staked ETH was slashed around the same time:

One validator slashed alone: close to zero.
1% of all stake slashed in the window: about 3% of each balance.
A third of all stake: the whole balance.

Pectra also raised the maximum effective balance to 2,048 ETH for compounding validators. One consolidated key can carry 64 times the stake of a 32 ETH validator, so one mistake slashes more ETH.

Liquid staking tokens inherit this risk; see liquid staking risks. Restaking adds a second set of slashing rules, covered in restaking risks.

Operational Causes in Practice

Three post-mortems show the pattern: one validator key, two live signers.

Date	Operator	Slashed	Cause
April 13, 2023	RockLogic (Lido)	11	Migrated keys came back on the old cluster
October 11, 2023	Launchnodes (Lido)	20	Failover sent two clients to one unprotected signer
September 10, 2025	Ankr cluster on SSV	39	Parallel instance run outside SSV during maintenance

RockLogic and Launchnodes paid 1 ETH per validator up front, under pre-Electra rules. The same mistake now costs less at first, but correlation still scales with the number of keys.

RockLogic: the migration that came back

After an execution client database corruption, RockLogic moved keys from one cluster to another. Prysm later re-imported the supposedly deleted keys during an update and restart.

Prysmatic Labs confirmed the bug and fixed it in v4.0.3. Lido's post-mortem recommends doppelganger protection and a "total wipe of initial cluster."

Launchnodes: a failover without a signing record

During datacenter connectivity problems, two validator clients, the original and a manual fallback, were pointed at one Web3Signer without slashing protection. Both signed, producing duplicate attestations.

Among the steps that would have prevented it, the post-mortem lists "Destroying the DC1 node cluster before failing over to DC2." It projects 28.677 ETH in penalties and missed rewards.

Ankr on SSV: maintenance outside the cluster

Ankr attributed it to "an internal maintenance mistake that ran a parallel validator instance outside of SSV." SSV says its protocol was not compromised.

SSV's lessons: keep keys in one environment, run one active instance, and protect slashing data during migrations and failovers.

Prevention Checklist

One active signer per key. No active-active failover. Put redundancy in DVT, not duplicate clients.
Slashing protection everywhere. Including remote signers such as Web3Signer.
Move the history with the key. Export and import the EIP-3076 interchange file on every migration.
Destroy before you start. Stop and wipe the old instance, confirm it is gone, then start the new one.
Turn on doppelganger detection. Lighthouse's stays silent for 2 to 3 epochs to listen for a duplicate.
Treat it as a last line. Lighthouse warns it is "imperfect" and that two instances are never safe.
Use DVT for uptime. Obol and SSV split a key across nodes that sign by threshold, such as 3 of 4.
Script and review migrations. Rehearse them on a testnet first.
Size keys deliberately. A 2,048 ETH validator puts 64 times the stake behind one error.
Spread operators. Correlation penalties grow when many keys fail together.

If you stake through an operator, ask for its migration runbook, signer setup and past post-mortems before you delegate.

How Guardian Helps

For institutions planning staking, Guardian's advisory covers custody and signing design, and assesses the counterparties you would depend on.

Guardian's infrastructure pentests test cloud posture, IAM, secrets handling, build and deploy, containers and clusters: the systems that run signing keys.

For keys in general, see our guide to private key management.

Frequently Asked Questions

What gets an Ethereum validator slashed?

Signing two different blocks for the same slot, two attestations with the same target epoch, or an attestation that surrounds another. Going offline is not slashable; it only costs rewards.

How much ETH do you lose when slashed?

Since Electra, 1/4,096 of effective balance up front (0.0078125 ETH at 32 ETH). About 18 days later comes a correlation penalty: tiny for an isolated slashing, the full balance if a third of all stake is slashed in the window.

How long does a slashed validator take to exit?

It is forced to exit and cannot withdraw for at least 8,192 epochs, about 36 days. While it waits, it is penalized for the attestation duties it no longer performs.

Does doppelganger protection prevent slashing?

It lowers the risk but does not remove it. Lighthouse calls it best-effort and imperfect: network faults can hide a duplicate, and it never makes running two instances of one key safe.

Sources

Pressure-test the systems behind your validator keys

Guardian advises on signing design and pentests the cloud, secrets and pipelines that decide whether one key ever signs from two places.

Talk to Guardian Explore institutional security