EigenLayer crossed $15 billion in total value locked earlier this year, making restaking one of the fastest-adopted primitives in Ethereum’s history. The pitch is straightforward: validators who already stake ETH can opt in to secure additional protocols - called Actively Validated Services, or AVSs - and earn extra yield in return. Shared security, recycled capital, everyone wins.
Except the risk accounting is murkier than that framing suggests.
What Restaking Actually Changes
In standard proof-of-stake, a validator’s staked ETH is at risk of being slashed if they behave maliciously or fail to meet liveness requirements on one network - Ethereum mainnet. Restaking changes that constraint. An operator who opts into multiple AVSs is now subject to slashing conditions across each of them simultaneously. Their stake isn’t multiplied; their exposure is.
This matters because AVS slashing conditions aren’t standardised. Each protocol defines its own criteria - what counts as a fault, how severe the penalty is, what the dispute resolution mechanism looks like. A validator running five AVSs is navigating five different threat surfaces, some of which may be immature codebases with undiscovered bugs in their slashing logic.
EigenLayer does have a slashing veto mechanism controlled by a multisig during its early phases, acting as a backstop against buggy or malicious slashing. But that’s a training wheel, not a permanent architecture.
The Correlation Problem
The deeper structural issue is correlated risk. If a widely-adopted AVS has a critical bug that triggers mass slashing across thousands of operators simultaneously, the downstream effect hits Ethereum’s validator set directly. This isn’t speculative - it’s the same category of risk that made analysts nervous about liquid staking derivatives becoming too large a share of Ethereum’s validator base.
Restaking doesn’t create systemic risk from nothing. But it increases the surface area over which a single bug, exploit, or flawed slashing parameter can propagate.
Yield Is Doing the Selling
Operators are opting in because the additional yield is real and immediate, while the tail risk of correlated slashing events remains theoretical. That asymmetry tends to drive adoption faster than prudence recommends.
Several AVSs are now live on EigenLayer mainnet, including EigenDA - Ethereum’s own data availability layer built on restaking. The model is already embedded in Ethereum’s infrastructure stack. The question of whether the risk has been properly distributed, or simply obscured by yield, hasn’t been settled by the market or by researchers.