Bitcoin L2 Security: Trust Models, Bridges and Exits

By Owen Thurm Published November 6, 2026 6 min read

Quick answer. A Bitcoin L2 is as secure as whoever can release its bitcoin, and many still trust a signer majority. Judge one by who holds the BTC, whether you can exit alone, and who can upgrade the bridge.

Watch: Bitcoin L2 Security: Trust Models, Bridges and Exits 1:39

Video transcript

A Bitcoin L2 is only as secure as whoever can release its bitcoin. And many still trust a signer majority.

The difference is real money. In September 2026, Liquid paid out 3,996 bitcoin against about 4,000 L-BTC created by a software bug, The Block reported.

Every design answers one question: who can release the BTC? Bitcoin Layers rates signer-run pegs like Liquid, Rootstock and Stacks High or Very High custody risk. Lightning, where users can exit alone, rates Low.

BitVM bridges cut trust but do not remove it. BitVM2 needs one honest setup committee member, one rational operator, and anyone willing to challenge. Citrea's Clementine adds an honest watchtower.

Many designs have a second spending path. Liquid has offline emergency keys held by Blockstream. Rootstock has a three of four emergency multisig. BOB's admin can upgrade the bridge with no exit window.

Score each layer on six questions. Who can release the BTC? Can you withdraw without permission? Where is the state data? Who runs the network? What finalizes state? And who can change the rules?

So score the weakest link. A strong peg does not help if the bridge contract above it can be upgraded overnight.

Read the full guide for the due diligence questions, or talk to Guardian about risk ratings that track who can upgrade, pause or drain your position.

"Bitcoin L2" covers very different systems. Some let you withdraw on your own. Others ask you to trust a federation, a custodian or an admin key.

The difference is real money. In September 2026, the Liquid Federation paid out 3,996 BTC against about 4,000 L-BTC that a software bug had created, The Block reported.

This guide sorts the main designs by what each trusts, then gives a scoring framework and due diligence questions.

Types of Bitcoin L2

Bitcoin cannot check another chain's state by itself. So every design answers one question differently: who can release the BTC?

Design	Example	Who can release the BTC	Bitcoin Layers custody risk
Federated sidechain	Liquid	11 of 15 watchmen	Very High
Merge-mined sidechain	Rootstock	5 of 9 PowPeg signers, keys in HSMs	High
Chain with a signer bridge	Stacks (sBTC)	14 signers, per Bitcoin Layers	High
Rollup with a BitVM2 bridge	Citrea	Operators, checked by any challenger	Under review
Rollup settling on Ethereum	BOB	Wrapped BTC custodians, plus the bridge admin	Very High
Payment channels	Lightning	Each user, by closing the channel	Low
Statechain-style transfers	Spark	The user and an operator, 2-of-2	Low

Bitcoin Layers rates each project on custody, data availability, operators and finality. It says its review "is not a security audit." We read these ratings on November 6, 2026.

For BTC wrapped on Ethereum, see our guide to wrapped Bitcoin risks.

Bridge Trust Models

Federations and multisigs

Liquid's peg is an 11-of-15 watchman multisig, per Blockstream's 2020 Liquid paper.

Its docs add two safeguards. Peg-ins wait 102 Bitcoin confirmations, and watchmen pay only users on a Peg-out Authorization Key list, which takes three days to change.

Rootstock's PowPeg is a 5-of-9 multisig with keys inside PowHSM devices. Rootstock says the devices sign only after enough Bitcoin proof of work.

Rootstock argues colluding signers cannot steal. Bitcoin Layers still rates its custody High risk, because users trust the PowPeg operators.

Stacks' sBTC sits in a taproot multisig held by its signers. Stacks launched it with 15; Bitcoin Layers now counts 14 and says 10 colluding could take all the BTC.

Signers are not the only thing to trust. Blockstream traced September's unbacked L-BTC to "a bug in the Elements software," SideSwap said, per The Block.

The peg-out used SideSwap's valid authorization key, which was not compromised. Our biggest bridge hacks ranking covers it.

BitVM bridges

BitVM2 checks a SNARK proof on Bitcoin optimistically. Operators are assumed honest unless a challenger proves fraud.

Its authors say the BitVM Bridge needs:

One honest setup committee member (1-of-n), not an honest majority.
One active, rational operator, for liveness.
Any user willing to challenge.

Withdrawals work by reimbursement. An operator fronts the user's BTC, then claims it back after a challenge period.

Bitcoin has no covenants to enforce that. So an n-of-n committee pre-signs the transactions, and safety rests on one member deleting their key.

Citrea's mainnet went live on January 27, 2026, issuing cBTC through Clementine, a bridge built on BitVM2.

Its docs say it needs one honest signer, one honest watchtower and one rational challenger. Reimbursement waits about 1.5 days for challenges.

The docs note two ways the deployed version differs from the whitepaper: challengers fund challenge costs upfront, and one watchtower check is omitted.

Unilateral exit

Lightning funds sit in a 2-of-2 channel multisig. Either party can close alone by broadcasting the latest state.

The catch is watching. If a counterparty broadcasts an old state and nobody contests it, they close with stale balances, Bitcoin Layers notes.

Spark pairs each user with an operator in a 2-of-2, with a pre-signed exit. Users trust the operator to delete old key shares after each transfer.

Hidden Multisigs and Upgrade Keys

Many designs have a second spending path. It guards against lost keys, and it is also a key.

Layer	Fallback	Held by
Liquid	Emergency keys, once coins sit unmoved 4,032 blocks (about 28 days)	Blockstream, offline; three keys
Rootstock	3-of-4 emergency multisig, after one year of inactivity	RootstockLabs, MoneyOnChain, Jameson Lopp, Adrian Eidelman
BOB	Bridge upgrade with no exit window	A centralized admin
BitVM2 bridges	Pre-signed spending paths; fixed operator set	The setup committee

Sources: Blockstream's Liquid paper and docs, Rootstock, Bitcoin Layers (BOB) and the BitVM2 paper.

For each fallback, ask who holds it, after what delay, and whether users can exit first.

Ethereum rollups have the same problem, as our ZK vs optimistic rollup comparison shows: upgrade keys outrank the proofs.

A Scoring Framework

Score each layer on six questions. The first five map to Bitcoin Layers' four categories; the last adds upgrade and emergency keys.

Question	Strongest answer	Weakest answer
Who can release the BTC?	You, with a Bitcoin transaction	One custodian, or undisclosed signers
Can you withdraw without permission?	Yes, unilateral exit	Only if signers approve
Where is the state data?	On Bitcoin, or held by you	An offchain committee
Who runs the network?	Anyone, peer to peer	One sequencer or a federation
What finalizes state?	Data posted to Bitcoin	The operators' signatures
Who can change the rules?	Nobody, or only after an exit window	An admin, instantly

Bitcoin Layers rates custody green only if "users can unilaterally exit the protocol with a Bitcoin L1 transaction."

Red needs at least five publicly known signers outside the team building the network. Anything less is rated "Stop!"

Score the weakest link. A strong peg does not help if the bridge contract above it can be upgraded overnight.

Due Diligence Questions

Who holds the BTC, at what threshold, and are the signers named?
Can I withdraw to Bitcoin without anyone's approval, and how long does it take?
Is there an emergency key? Who holds it, after what timelock?
Who can upgrade the bridge contracts, and do users get an exit window first?
For BitVM bridges: who sits on the committee, who operates, and can anyone challenge?
Which software mints the wrapped BTC, and who reviewed it?
Is the BTC I hold this layer's own peg or a third-party wrapper?
Can I reconcile BTC held against tokens issued, onchain?

How Guardian Helps

For allocators, Guardian's risk ratings review "who can upgrade, pause, or drain" a position. They cover signer sets, thresholds, timelocks and admin keys.

Guardian keeps watching signers, thresholds, timelocks and bridge routes. Material changes reach you the day they happen.

Guardian's public portfolio includes reviews of Citrea's Stablecoin Bridge (September 2025) and token (March 2026).

For protocol teams, Managed Security monitors contracts on every block: privileged calls, role changes, upgrades and multisig transactions.

Frequently Asked Questions

What is a Bitcoin L2?

A system that moves bitcoin activity off the base chain: a sidechain, rollup, payment channel or similar design. The term is loose. Many Bitcoin L2s are separate chains whose BTC is held by a federation, not secured by Bitcoin itself.

Which Bitcoin L2 design is the most trust-minimized?

Designs with unilateral exit, such as Lightning channels, let users withdraw with a Bitcoin transaction alone, and Bitcoin Layers rates their custody Low risk. The federated and signer-run pegs reviewed here rate High or Very High.

Is a BitVM bridge trustless?

No. BitVM2's bridge design needs one honest member of its setup committee and one active, rational operator, though anyone can challenge fraud. Citrea's Clementine adds one honest watchtower, and its docs note two departures from the whitepaper.

What happened to Liquid in September 2026?

About 4,000 L-BTC, which Blockstream traced to a bug in the Elements software, were pegged out for 3,996 BTC, The Block reported. Liquid paused, Blockstream patched the bridge nodes, and the attacker returned 3,400 BTC.

Sources

Know who can release the BTC you hold

Guardian underwrites who controls a position and the bridges it depends on. Material changes reach you the day they happen.

Talk to Guardian