
Whether a rollup calls itself “optimistic” or “ZK” tells you how it proves correctness. It tells you almost nothing about who can still unilaterally change its contracts, pause withdrawals, or override its own proof system in an emergency — and in most rollups running today, at least one party retains exactly that kind of power. A maturity scale for exactly this question, commonly known as the stages framework, gives a more complete way to evaluate an L2’s actual security than its proof type alone.
The stages framework: Stage 0 through Stage 2
The framework scores decentralization and user-exit guarantees, not bug-freedom — a project can satisfy every requirement at a given stage and still contain an undiscovered flaw in its code. With that caveat, the three stages work roughly as follows:
Stage 0
The baseline: the project must self-identify as a rollup, post its state roots and the data needed to reconstruct that state on L1, and run a proper proof system — one that can reject an invalid root (for fraud-proof designs) or that only accepts a root backed by a valid proof (for validity-proof designs). At least five parties outside the core team must be able to actually submit a fraud proof, so the check is not purely theoretical even at this earliest stage.
Stage 1
Governance over upgrades moves into smart contracts rather than a single team’s direct control, and a Security Council — a defined multisig, required to have at least eight members with a 75% approval threshold, with public identities or pseudonyms — can intervene on bugs. Crucially, any upgrade initiated outside the Security Council must give users at least a 5-day exit window before it takes effect, and optimistic rollups’ challenge periods must be at least 5 days. The Security Council’s own power to block or force L2-to-L1 messages requires compromising at least 75% of its own membership, aside from genuine bug response — it is bounded, not absolute.
Stage 2
Fraud proof submission becomes fully permissionless — anyone can submit one, not just an allowlisted set of parties. Users must get at least a 30-day exit window before any unwanted upgrade takes effect, including ones initiated by the project’s own governance process. The Security Council’s remaining power narrows further, limited to acting promptly only on bugs that are objectively verifiable on-chain, such as two contradicting valid proofs for the same state.
Very few rollups in production today sit at Stage 2. Most sit at Stage 0 or Stage 1, which means a training-wheels-off, fully permissionless, long-exit-window system is still more the destination than the present reality for the majority of the ecosystem — a detail that matters more than a project’s own marketing language about being “decentralized.”
Escape hatches: the user’s last resort
An escape hatch is a mechanism — usually a smart contract on L1 — that lets users withdraw their funds directly, bypassing the rollup’s normal operator or sequencer entirely, if that operator becomes unresponsive, malicious, or is shut down. Whether an escape hatch exists, whether it has ever actually been used under real conditions rather than only described in documentation, and how long and costly it is to use, are concrete, checkable facts about a specific rollup — not a theoretical property every rollup automatically has just by being a rollup.
Reading a specific rollup’s security model
Four concrete questions cover most of what matters: what proof system does it use, and is that a fraud-proof or validity-proof design (see the companion guide on rollup types); what stage does it currently sit at on an independent tracker, not just its own claim; what can its Security Council or upgrade key do unilaterally, and under what threshold; and does a working, exercised escape hatch exist for users if the operator disappears. None of these require trusting the project’s own self-description — all four are checkable against independent sources and the project’s own published contract addresses and governance documentation.
Why a framework like this exists at all
Early rollups shipped with a single team controlling an upgrade key, a short or nonexistent public challenge process, and no meaningful way for an ordinary user to exit if that team acted against their interests — a reasonable, deliberate shortcut to get a working system live quickly, not a secret design flaw. As the ecosystem matured, the gap between “a rollup” in the loose marketing sense and “a rollup with the security guarantees people assume a rollup has” became wide enough that an independent, checkable classification became genuinely useful, rather than relying on every project’s own account of its own decentralization.
The Security Council is a feature, not automatically a flaw
It is tempting to read any centralized emergency power as inherently bad, but the framework treats a well-constituted Security Council as a legitimate, bounded safety mechanism — a way to respond to a genuine on-chain bug faster than a fully permissionless process could, while still constraining what that council can do outside of bug response with a high membership threshold and a public roster. The actual question worth asking about any specific rollup’s council is not whether it exists, but how tightly its power is bounded and how independent its members actually are from the core team and from each other.
A rollup can also move backward as well as forward on this scale if, for example, its fraud-proof system is temporarily disabled during an upgrade — stage classification describes a point in time, not a permanent certification, which is one more reason to check a current, independently maintained source rather than relying on a project’s own historical claim.
Treat the stage label the way you would treat any other point-in-time disclosure: useful, specific, and worth re-checking before relying on it for anything that actually matters to you.
A reasonable habit: before bridging meaningful value into any specific rollup, spend five minutes checking its current stage and council powers on an independent tracker rather than taking its own homepage copy at face value.
That five-minute check is cheap insurance against relying on a stage claim that changed since the last time anyone looked.
It also travels well: the same five questions apply whether the rollup in question is optimistic or ZK, new or long-established, since they are about governance and exit rights rather than proof mechanics.
In short: read the stage, read the council, and read the escape hatch — in that order — before assuming a rollup’s security matches whatever its name or its proof type implies.
None of this requires specialized tooling — a project’s own documentation, cross-checked against an independent tracker, is usually enough.
Risk note
A rollup’s stage and its proof type are two different, independently checkable facts, and a low stage is not a prediction of failure — it is a statement about how much is currently resting on trust in a centralized party rather than on permissionless, code-enforced guarantees. Treat stage classification as a disclosure to read, not a score to chase.
Not financial advice. Nothing on this page is a recommendation to buy, sell, stake, bridge, or avoid any asset or protocol.
FAQ
Does Stage 2 mean a rollup is bug-free?
No. The stages framework scores decentralization and user-exit guarantees, not code correctness. A Stage 2 rollup can still contain an undiscovered flaw; the framework’s own documentation says explicitly that it does not assess bug risk.
What’s the difference between a rollup’s proof type and its stage?
Proof type (optimistic or ZK) describes how correctness is verified. Stage describes who controls upgrades, how long users have to exit before a change takes effect, and whether fraud proof submission is open to anyone. The two are independent; a rollup can pair either proof type with any stage.
What is an escape hatch?
A mechanism, usually an L1 smart contract, that lets users withdraw funds directly if the rollup’s normal operator becomes unresponsive or malicious, bypassing the usual process entirely.
Are most rollups at Stage 2 today?
No. Most production rollups currently sit at Stage 0 or Stage 1, meaning a centralized party still retains meaningful power over upgrades or proof submission for the large majority of the ecosystem.
Frequently asked questions
Does Stage 2 mean a rollup is bug-free?
What's the difference between a rollup's proof type and its stage?
What is an escape hatch?
Are most rollups at Stage 2 today?
Educational content, human-edited and possibly AI-assisted. It describes how systems work and is not financial, legal or tax advice. Report an error via our corrections page.


