Glossary · Proof of stake

Slashing condition

Published 24 May 2025 · Updated 20 Oct 2026 · 11 min read

A slashing condition is a rule in a proof-of-stake protocol that defines provable validator misbehavior, such as signing two conflicting blocks or votes. When the evidence reaches the chain, the protocol takes part of the validator's stake. On Ethereum it also forces the validator out, while a missed duty alone is not slashable.

  • A slashing condition targets provable misbehavior, most often two conflicting messages signed with one key.
  • Since Pectra, an isolated Ethereum slashing costs about 0.18% of stake, while a mass event can take all of it.
  • Ethereum slashings with a published cause came from operational failures, such as one key on two machines.

How slashing conditions work

In proof of stake, validators commit staking deposits to support a chain's consensus rules. Rewards alone create the nothing-at-stake problem, where validators can support competing forks without risking anything. A slashing condition adds a cost to contradictory commitments. Anyone can verify the signatures and rule violation, without trusting the reporter.

Number 1

Sign

Each validator in proof of stake signs blocks and votes with its key. These signatures identify the signer and make its messages checkable.

Number 2

Prove

Anyone can submit two conflicting signed messages as evidence. A block proposer includes them, letting the protocol verify the violation.

Number 3

Penalize

The protocol burns or confiscates stake once the evidence meets a slashing condition. Ethereum starts with an initial penalty, then applies further ones.

Number 4

Remove

Ethereum forces the validator to exit, delaying access to its remaining balance. Cosmos Hub uses jailing, with permanent tombstoning after double signing.

Ethereum's three slashable offenses

Ethereum's beacon chain separates block proposals from votes, called attestations. An Ethereum attestation has source and target checkpoints, identified by epochs. Each Ethereum epoch contains 32 slots and lasts 6.4 minutes. Exactly three acts meet a slashing condition.

Two blocks, one slot

A proposer signs two different blocks for one Ethereum slot, which lasts 12 seconds. The conflicting proposals undermine agreement on which chain should become final.

Double vote

A validator signs two different attestations with the same target epoch. The validator contradicts its own vote for that epoch, breaking the voting rules that protect finality.

Surround vote

A validator signs an attestation whose source is earlier and target is later than another vote's corresponding epochs. Strictly surrounding its other vote breaks Casper FFG's finality rules.

Figure 1 · Casper FFG rulesCompare two validator votesChoose a scenario or move vote B's endpoints to see whether the vote pair is slashable.
Scenario

Each arc runs from a vote's source epoch to its target epoch.

Vote A
Sourceepoch 101Targetepoch 104
Vote B
Sourceepoch 104Targetepoch 105
Not slashable: separate spansOne vote ends where or before the other starts. Neither vote surrounds the other.no shared target, no surround → not slashable

This is simplified. Real votes name blocks. Two votes for different blocks with identical epochs constitute a double vote.

Missing duties does not trigger Ethereum slashing. A missed attestation incurs source and target penalties, but missing a head vote or block proposal carries no penalty. The inactivity leak begins after more than four epochs without finality. It reduces inactive validators' stake until active validators can finalize again. This is a separate penalty, not a slashing condition.

What Ethereum slashing costs

Since Pectra (May 2025), day 0 burns 1/4096 of effective balance and starts the forced exit. Source and target attestation penalties continue for 36.4 days because its votes no longer count. On day 18.2, the correlation penalty is roughly effective balance times three times the share of all staked ETH slashed in the window. That window covers about 18 days before and after slashing, and the penalty cannot exceed the whole balance. On day 36.4, the remainder becomes withdrawable.

Pectra's EIP-7251, activated on 7 May 2025, lowered the initial penalty. It also let consolidated validators hold up to 2,048 ETH of effective balance. In the October 2026 calculation, an isolated 32 ETH slashing costs about 0.058 ETH in penalties, or 0.18%, mostly from missed duties. Before Pectra, this case cost about 1.05 ETH, roughly 18 times more. If 1% of all stake is slashed in the window, the loss reaches about 1.02 ETH. At one third or more, the whole balance is lost.

Figure 2 · Ethereum consensus specsCompare Ethereum penalty scenariosChoose the rules, validator size and stake slashed together. See each penalty, the total loss and when deductions happen.
Rules
Effective balance
Slashed in the same window
Initial penalty0.0078125 ETHDay 0, burned at once
Missed duties0.04997 ETHEvery epoch until day 36
Correlation penalty0.000069888 ETHDay 18
Total loss0.05785 ETH0.18% of the balance
Day 0Slashed. The initial penalty is burned and the exit starts.Day 18.2The correlation penalty is applied.Day 36.4Withdrawable. The sweep pays out what is left.
Initial penalty: 32 ETH / 4,096 = 0.0078125 ETHCorrelation penalty: 32 ETH × 3 × 32 ETH / 43,948,738 ETH = 0.000069888 ETHProposer reward for including the proof: 0.0078125 ETH

Calculated from consensus specs with ethereum.org's 20 October 2026 total stake, a short exit queue and no inactivity leak. Missed rewards are excluded.

Why validators get slashed in practice

Rated Explorer records 594 slashed validator keys across 152 events in its 20 October 2026 reading. Slashed ETH represents 0.001% of staked ETH. The incidents with published causes were operational failures, not attacks on Ethereum's consensus rules.

  1. February 2021

    Staked disables protection

    Staked had 75 validators slashed after disabling its Prysm client's protection database. The penalty was about 18 ETH, with compensation promised.

  2. April 2023

    RockLogic failover duplicates keys

    RockLogic had 11 validators slashed after failover left identical keys active in two clusters. A client bug also re-imported deleted keys.

  3. October 2023

    Launchnodes fallback lacks protection

    Launchnodes had 20 validators slashed during a data center fallback. Its clients shared one Web3Signer instance without slashing protection enabled.

  4. September 2025

    Ankr maintenance starts a parallel instance

    On 10 September 2025, SSV Network clusters saw one validator and then 39 Ankr validators slashed. Ankr confirmed a maintenance mistake that ran a parallel instance outside SSV.

  5. March 2026

    Lido CSM operator slashed

    Six Lido Community Staking Module validators were slashed on 12 March 2026. Penalties and missed rewards totaled about 0.8 ETH. The cause remains unpublished in October 2026.

How operators avoid slashing

A slashing condition can be met during routine maintenance, even without an attack. Lido's operator guide says to wait at least one hour and check that keys are deleted before importing them on a new machine. The slashing protection database must move with them.

Slashing protection database

Clients record blocks and attestations they sign, then refuse conflicting messages. EIP-3076 standardizes signing history for transfers, so it can travel with the keys.

Doppelganger protection

After restarting, clients listen for their keys before signing. Lighthouse, Teku, Prysm and Nimbus support this protection, with Nimbus enabling it by default. Client docs call it best-effort, not a way to run two nodes with one key.

Remote signer

Web3Signer keeps keys outside validator clients and maintains its own protection database. Multiple signer instances can share that database rather than maintain separate histories.

Distributed validator technology (DVT)

Obol and SSV split keys into operator-held shares without rebuilding the full key. Signing requires agreement from a threshold, such as three of four operators.

Figure 3 · Simplified failoverTest a validator failoverToggle doppelganger protection and follow the backup's response. Compare double signing with a restart that detects the active key and stops.
Validator node iconPrimary nodeSigns one vote per epoch
Validator node iconBackup nodeOff. The key runs on one machine
Block iconBeacon chainCounts the primary's votes
Signed messagesEpoch NPrimary signs its vote
Step 1 of 4 · Normal dutyThe primary signs its expected attestations each epoch. The backup stays inactive and does not use the key.

Simplified scenario with example epochs. Lighthouse waits 2 to 3 epochs, about 12 to 20 minutes, before signing.

Slashing across proof-of-stake networks

Penalty rules and exposure differ across proof-of-stake networks. The comparison reflects October 2026 rules, not a shared slashing condition. Jailing suspends participation, while Cosmos Hub's tombstoning permanently excludes a validator after double signing.

NetworkWhat is slashedDowntimeWhose stake is at risk
EthereumWhat is slashedConflicting signatures burn 1/4096 of effective balance initially, with total losses potentially reaching 100%.DowntimeMissed attestations incur source and target penalties, without slashing.Whose stake is at riskOnly the slashed validator's balance, with no protocol-level delegation.
Cosmos HubWhat is slashedDouble signing slashes 5% of stake and permanently tombstones the validator.DowntimeMissing over 9,500 of the last 10,000 blocks costs 0.01% and ten-minute jailing.Whose stake is at riskValidators and their delegators share the loss.
PolkadotWhat is slashedThe equivocation penalty is (3 × offenders ÷ active validators)² of stake, capped at 100%.DowntimeIsolated unresponsiveness has the lowest severity, with removal or a slash up to 0.1%.Whose stake is at riskSince July 2026, validators risk their stake, but nominators do not.
TezosWhat is slashedDouble baking costs 5%. Double attestation costs 9 × the offending stake share², capped at 100%.DowntimeDowntime is not slashed. Only double signing is.Whose stake is at riskBakers and stakers lose stake. Plain delegations are not slashed.
SolanaWhat is slashedNo protocol slashing exists. SIMD-0204 proposes logging evidence without reducing stake.DowntimeThere is no protocol slashing for downtime.Whose stake is at riskNo stake faces a protocol slash.
CardanoWhat is slashedNothing is slashed. Cardano has no slashing.DowntimeAn underperforming pool misses staking rewards.Whose stake is at riskDelegated ADA is never at risk.
AvalancheWhat is slashedNothing is slashed. Avalanche has no slashing.DowntimeValidations started since Helicon (September 2026) lose rewards below 90% uptime.Whose stake is at riskValidator stake is not exposed to slashing.

In October 2026, NEAR and Aptos have not implemented slashing. Sui cuts the staking rewards of low-scored validators for an epoch. Its docs describe cuts to rewards, not to stake. Delegators need network-specific rules to understand whether poor performance threatens rewards or deposited stake.

Slashing in restaking and liquid staking

Additional services can introduce a slashing condition beyond a chain's base rules. Restaking puts assets behind those services, while liquid staking determines how validator losses reach token holders. Bitcoin staking can also make conflicting signatures a reason to destroy stake. Symbiotic's full mainnet, with configurable vault slashing, went live on 28 January 2025.

EigenLayer

Restaked ETH for services

Services called AVSs define their own slashing conditions, and operators choose which operator sets to join. Unique Stake limits a given stake to one service. Slashed funds are burned unless the operator set was created as redistributable. Since July 2025, those sets can send them to a chosen recipient.

Mainnet slashing activated on 17 April 2025.

Babylon

Bitcoin staking for proof-of-stake chains

Finality providers use extractable one-time signatures, or EOTS. Signing two different blocks at one height exposes the provider's private key. Anyone can then complete pre-signed Bitcoin transactions that burn part of the stake. Babylon Genesis launched on 10 April 2025.

Live slashing rate of 0.1% in October 2026.

Lido

Shared exposure through stETH

Uncovered validator losses fall on all stETH holders. The Reserve Fund could cover some estimated risks, with decisions made by the Lido DAO. Community Staking Module operators post bonds that absorb penalties first.

Reserve Fund cover is decided by a Lido DAO vote.

Rocket Pool

Operator bonds absorb penalties

Node operators post an ETH bond for each validator, and penalties come from that bond first. The minimum is 4 ETH following Saturn 1, down from the previous 8 ETH requirement.

Saturn 1 lowered bonds on 18 February 2026.

Building staking products around slashing rules

Each network's slashing conditions should shape a staking product's signing controls, accounting and user information. We build staking logic, reward distribution and user dashboards through our staking platform development service. The product specification should also state who absorbs penalties and how operators handle key transfers.

PixelPlex case icon

We operate a validator on the Canton Network, which does not slash stake, and we built CC View, our Canton explorer. Its validator health views let operators monitor liveness, version health, traffic and rewards. For Mantra DAO, we deployed 60+ staking pools on Ethereum, BNB Smart Chain and Polygon, with $23M staked in April 2022.

  1. Key custody and signing

    Define where validator keys live and which instances can request signatures. Keep a full signing key from running in parallel setups.
  2. Protection and failover

    Treat signing history as part of every key migration. Document how a backup starts, detects existing activity and avoids duplicate signing.
  3. Monitoring and alerts

    Track validator liveness and key activity during restarts. Alerts should distinguish ordinary missed duties from evidence of conflicting signatures.
  4. Contracts and reward accounting

    Separate rewards, initial deductions, later penalties and remaining balances in the product's accounting. Show users which losses fall on operators or stakers.
  5. Audits and risk disclosure

    Include smart contract audits in the build plan. Explain network-specific penalties, operational safeguards and any limits on coverage before users commit stake.

Discuss the safeguards your staking product needs

Share your network and operating model with us to discuss signing controls, penalty accounting and the scope of a staking product.

Discuss your project

Slashing condition FAQ

Can a validator be slashed for being offline?

No, not on Ethereum. While the chain finalizes normally, an offline 32 ETH validator loses about 0.0014 ETH a day, calculated from the consensus specs using October 2026 stake. On the Cosmos Hub, downtime is slashable at 0.01%, with 10 minutes in jail.

How much ETH does a slashed validator lose?

On its own, a 32 ETH validator gives up about 0.058 ETH in penalties, 0.18%, calculated from the consensus specs with October 2026 stake. A 2,048 ETH validator gives up about 3.98 ETH. Both also stop earning rewards until they exit. With 1% of all staked ETH slashed in the same window, the 32 ETH case grows to about 1.02 ETH.

How does slashing differ from the inactivity leak?

Slashing punishes provable signing violations, while the inactivity leak responds to a chain that has gone more than four epochs without finalizing. Inactive validators then keep losing stake until they control less than one third of the total, so the active validators can restore finality without them.

Can slashed funds ever be recovered or refunded?

Not from the protocol, because slashed ETH is burned. Operators and protocols can compensate holders from their own funds. Staked promised full compensation in 2021, and Lido covered RockLogic's losses from its own fund after a DAO vote in 2023. On Polkadot, governance has 27 days to cancel a slash before it applies.

Do liquid staking holders lose money after slashing?

They can. A Lido validator's slashing loss is shared across stETH holders unless something covers it. In the Community Staking Module, the operator's bond pays first, and the DAO can vote to use the Reserve Fund. After the March 2026 incident, about 0.35 ETH came from the operator's bond and the operator paid the rest.

Does Solana have protocol slashing for validators?

Solana has no protocol slashing. SIMD-0204, still under review, would log evidence of slashable behavior without reducing stake. A separate penalty proposal remains under discussion. Alpenglow, Solana's new consensus, is scheduled to activate on mainnet on 21 October 2026. It does not add stake slashing.

What makes an Ethereum attestation a surround vote?

An attestation is a surround vote when its source is earlier than another vote's source and its target is later than that vote's target, both signed by the same validator. Either vote in the pair can be the surrounding one. The slashing condition checks both orders.
Sources (49)

Get updates about blockchain, technologies and our company

We will process the personal data you provide in accordance with our Privacy policy. You can unsubscribe or change your preferences at any time by clicking the link in any email.

Follow us on social networks and don't miss the latest tech news

  • facebook
  • X
  • linkedin
  • instagram
Stay tuned and add value to your feed