EigenLayer’s core pitch is elegant on its surface: validators who already stake ETH to secure Ethereum can simultaneously extend that economic security to other protocols, called Actively Validated Services (AVSs). One validator, multiple commitments, shared collateral. The appeal is obvious - new protocols get cryptoeconomic security without bootstrapping a token and a validator set from scratch.
The problem is that “shared security” is doing enormous lifting in that sentence, and the risk math doesn’t actually work the way the marketing implies.
Slashing Conditions Don’t Stack Cleanly
When a validator opts into multiple AVSs through restaking, each service can theoretically slash their staked ETH for misbehaviour specific to that service. In theory, this creates aligned incentives. In practice, slashing conditions across AVSs are heterogeneous - different services define misbehaviour differently, enforce it through different mechanisms, and have different latency requirements for dispute resolution.
A validator running infrastructure for three AVSs simultaneously is operating under three distinct threat models. If two of those models conflict - say, one AVS requires a fast response that another penalises as a double-sign - the validator is in an adversarial position that the shared collateral model didn’t anticipate. EigenLayer’s architecture places the burden of AVS design on individual service developers, which means the security quality a restaker provides varies significantly depending on which services they’ve opted into.
The “Security” Being Borrowed Is Correlated
Here’s the part that gets glossed over: the security restaking provides to AVSs isn’t independent security. It’s correlated to Ethereum’s validator set.
If a systemic shock hits ETH prices, restakers across all AVSs simultaneously become less well-capitalised. If there’s a coordinated slashing event on Ethereum mainnet - unlikely but not impossible - the collateral underpinning dozens of AVSs degrades at the same time. This is not how the word “security” typically works in finance or cryptography. You don’t improve a system’s resilience by making its guarantees contingent on the same underlying asset.
Operator Concentration Is Already Visible
Data from EigenLayer’s own dashboard has consistently shown that a small number of large operators - institutional staking providers - account for a disproportionate share of restaked ETH. This matters because AVSs that believe they’re buying distributed security may actually be buying exposure to three or four operators running highly similar infrastructure stacks.
Restaking isn’t fraudulent and the technical architecture is genuinely novel. But describing it as “borrowed security” is more accurate than “shared security.” What’s being borrowed hasn’t been fully stress-tested, and the lending terms are still being written.