A new paper from LZ Research argues that slashing, the core deterrent underpinning proof-of-stake security, may fail against rational validators who coordinate before misbehaving.

Researchers Hao Chung and Chen-Da Liu-Zhang uploaded the paper, titled "Too Late to Slash: Coordinating a Risk-Free Equivocation Attack," to arXiv on September 24, 2026, under identifier 2609.30509. Their central claim is that validators can organize an equivocation attack in a way that removes the financial risk slashing is supposed to impose.

In proof-of-stake systems, validators lock tokens as collateral for the right to help confirm transactions. Slashing is the automatic punishment: if a validator cheats, the protocol destroys some or all of its stake. The specific violation the paper examines is equivocation — a validator signing two conflicting blocks, effectively endorsing two incompatible versions of the chain's history at once.

Chung and Liu-Zhang propose a coordination protocol that lets an attacker solicit equivocation from validators without exposing them to losses if the effort falls short. If too few validators join, nobody signs the conflicting messages and nobody gets slashed. If enough participate, the attack goes forward. The downside case simply dissolves.

The authors frame the result in game-theoretic terms. They state that the coordination strategy produces an ex post Nash equilibrium — meaning the strategy remains optimal even after participants observe the outcome. They further argue the approach yields positive gains, expressed as ε > 0, regardless of how small those gains are relative to the amounts staked.

That claim cuts against the conventional security argument. Standard reasoning holds that an attack is rational only if the payoff exceeds the stake at risk. The paper contends that even a minimal successful attack comes out ahead, because coordination engineers the risk side of the equation down to zero.

The work is theoretical. It does not analyze any specific proof-of-stake protocol or token, and translating a game-theoretic construction into a live attack faces practical hurdles the model abstracts away. Validators would need to find one another, trust the coordination process, and act in concert — no trivial coordination problem in itself.

The stakes of the argument are structural. Proof-of-stake abandoned the physical cost that secures proof-of-work systems: miners burn electricity and hardware to produce blocks, making attacks inherently expensive. Validators mostly just sign messages, and signing a second, conflicting message costs almost nothing computationally. A 2023 report by a16z crypto emphasized slashing as the mechanism that raises the cost of corruption and collusion among validators, protecting network integrity against internal and external threats.

The paper's title states the critique plainly: by the time a protocol can punish anyone, a coordinated attack has either succeeded or quietly dissolved without consequence.

For protocol designers, the findings point toward incentive mechanisms that make coordination itself harder or riskier, rather than relying solely on penalties that activate after an attack becomes visible. For individual stakers, the practical read is narrower — slashing still applies to solo mistakes and uncoordinated misbehavior, and the open question is whether it holds against validators who plan together and commit only once the arithmetic favors them.

The paper is likely to draw scrutiny from protocol research teams at major proof-of-stake networks in the coming months, as they assess whether their own slashing parameters and coordination defenses withstand the ex post equilibrium the authors describe.