EigenLayer’s core pitch is elegant: validators who already stake ETH to secure Ethereum can pledge that same stake to secure other protocols - called Actively Validated Services - and collect additional yield in return. No new capital required. The stake does double duty.
The problem is that ‘double duty’ is doing a lot of work in that sentence.
What Restaking Actually Does
When a validator opts into an AVS through EigenLayer, they expose their staked ETH to a second set of slashing conditions. If they behave honestly on both Ethereum and the AVS, nothing changes. But if they violate the AVS’s rules - or if those rules are ambiguous, exploited, or poorly designed - they can lose ETH that was originally committed to Ethereum’s security.
The ETH doesn’t multiply. The obligations do.
Ethereum’s validator set is secured by roughly 33 million ETH at time of writing. That capital is credibly committed because the slashing conditions are narrow, well-understood, and governed by a protocol with years of adversarial testing. Restaking borrows the reputation of that commitment without fully inheriting the conditions that make it credible.
The Correlation Risk Problem
The deeper issue isn’t a single validator getting slashed across two systems. It’s what happens when many validators share the same restaked exposure. If a popular AVS has a bug or a governance failure that triggers mass slashing, the affected ETH comes directly out of Ethereum’s validator set. The two systems are no longer independent - they’re correlated.
This is the scenario that makes Ethereum core researchers cautious. Vitalik Buterin has written publicly about his concern that overloading Ethereum’s social consensus through restaking could force the protocol into positions where it’s expected to bail out AVS failures, blurring the line between Ethereum’s own security guarantees and those of third-party services built on top.
The Yield Isn’t Free
Restaking rewards exist because AVS protocols are paying for security they couldn’t otherwise afford. That’s rational. But validators accepting those rewards are implicitly underwriting risk that isn’t fully priced into their decision - particularly tail risk from AVS designs that haven’t faced real adversarial conditions yet.
EigenLayer has introduced slashing parameters and veto mechanisms to limit downside, but these are still relatively untested at scale. The architecture assumes that slashing events will be rare, isolated, and detectable in advance. High-stakes systems tend to fail precisely in the gaps between those assumptions.
Whether restaking matures into a stable layer of Ethereum’s infrastructure or becomes a source of correlated fragility depends almost entirely on how AVS design practices develop over the next few years - and that’s genuinely an open question.