Bitcoin L2s After the Liquid Hack: Where the Trust Lives

On 6 September 2026, an attacker drained roughly 4,000 BTC, about $320 million at the time, out of the Liquid Network in a little over half an hour. No private key was stolen. No federation member was bribed. The attacker found a bug in the code that Liquid uses to verify confidential transactions, minted almost 4,000 units of unbacked L-BTC, and pegged them out to real Bitcoin before anyone could react. Liquid’s fifteen-member federation, which normally authorizes withdrawals with an eleven-of-fifteen multisig, waved the transaction through, because a multisig cannot catch a consensus bug it does not know is there, according to the technical breakdown published by crypto.news and analyzed by Chainalysis.

The Liquid drain is the largest incident yet on a Bitcoin Layer 2, and it arrives at an awkward moment for the whole sector. Bitcoin itself trades near $77,300, roughly 39 percent below its October 2025 record of $126,080, with a market capitalization around $1.55 trillion (CoinGecko). The amount of BTC locked into second layers and Bitcoin DeFi has fallen much harder than the price, and one of the most polished projects in the field shut itself down in June. The hack broke no cryptography. It broke an assumption: that moving your Bitcoin onto a faster, cheaper, more programmable layer does not change who can take it from you. This guide walks the entire Bitcoin L2 landscape through that single lens, because the Liquid incident is the clearest illustration in years of where the trust in these systems actually sits.

A $320 Million Reminder of Where Trust Lives

Most of the big crypto losses of 2026 have followed a familiar script: a compromised signing key, a phished operator, a rogue insider with too much access. Security firms spent the year repeating that keys and admin permissions, not clever code, are where the money leaks out. Liquid inverted that story. The federation’s keys were never touched, and Blockstream, which builds the network, confirmed that no functionary was compromised. The attacker instead abused a caching shortcut deep inside Elements, the software that runs Liquid, and turned a performance optimization into a mint button.

That distinction matters because it changes the lesson. A stolen key is a failure of custody, and custody can be hardened with better hardware and stricter process. A consensus bug is a failure of the code that every participant trusts by default, and no amount of key hygiene protects against it. When the range-proof check misfired, the eleven honest functionaries who signed the fraudulent peg-out were not careless; they were doing exactly what the protocol told them to do, verifying a proof their own software had already marked valid.

By 10 September, Liquid had restarted block production on version 23.3.4 of Elements, though it was still processing empty blocks while functionaries checked the network’s integrity, and peg-ins and peg-outs remained frozen (The Crypto Times). Blockstream chief executive Adam Back said publicly that the one-to-one L-BTC to BTC peg would be covered, and urged holders not to panic-sell their L-BTC over the counter while the reserve was being restored. The attacker, who left on-chain messages claiming to be a white hat, returned 3,400 BTC (around $270 million) but kept 598.5 BTC, roughly $47 million, as a self-declared bounty; Blockstream called it theft and refused to pay.

Why Bitcoin Needs a Second Layer

Bitcoin’s base layer is deliberately slow. Blocks arrive roughly every ten minutes and carry only a few megabytes of data, which caps the network at a handful of transactions per second worldwide. That constraint is a feature, not a bug: it keeps a full node cheap enough to run on a home computer, which is what lets ordinary users verify the chain themselves instead of trusting a company to do it for them. The blocksize war of 2015 to 2017, which ended with the community rejecting larger blocks and adopting Segregated Witness, settled the question for a generation. Bitcoin would stay small at the base and push scale upward, onto other layers.

A second layer, in the cleanest definition, is a system that lets people transact in bitcoin without writing every payment to the main chain, while still being able to fall back to that chain if the layer misbehaves. The prize is obvious: instant, cheap payments; smart contracts; tokens and stablecoins; lending and trading, all denominated in the hardest money in crypto. The catch is equally obvious once you have watched the Liquid drain. Every one of these systems has to answer the same question, which is who holds the real bitcoin while you hold the layer’s version of it, and what happens if that party fails, is hacked, or simply refuses to give it back.

Ethereum solved a version of this problem with rollups that inherit the base chain’s security and let users exit unilaterally. Bitcoin cannot copy that design directly, because its scripting language is intentionally limited and cannot verify the state of another chain the way a general-purpose smart contract can. So the Bitcoin L2 field has splintered into a dozen different architectures, each making a different trade between speed, programmability, and trust. Sorting them is less about branding than about one practical test.

What Counts as a Bitcoin L2: The Exit Test

The most useful yardstick in this field is not throughput or total value locked. It is the exit test, articulated by the independent researchers at Bitcoin Layers, whose framework rates each system by how much its users depend on custodians, data availability, operators, and finality. The strict version of the test asks a single question: if the operators of the layer turned hostile or vanished tomorrow, could you recover your bitcoin to the base chain on your own, without their permission? If the answer is yes, the layer only borrowed your trust. If the answer is no, it took it.

This is not an academic distinction. Ethereum co-founder Vitalik Buterin warned back in January 2022 that there are fundamental limits to the security of bridges that move assets across multiple zones of sovereignty, and that the more value flows across such a bridge, the stronger the incentive to attack it (Cointelegraph). Liquid proved his point in the most expensive way available. When the peg broke, L-BTC holders had no unilateral escape hatch; they could only wait for the federation to restore the reserve and honor the peg. A trustless layer would not have put them in that position.

Grouping the field by the exit test rather than by marketing gives a spectrum that runs from trustless at one end to fully custodial at the other. The table below maps the main categories. Read it as a trust ladder: each rung down trades a little more of Bitcoin’s self-custody promise for a little more speed or programmability.

Anatomy of the Liquid Drain

Liquid uses confidential transactions, which hide the amounts being moved behind cryptographic commitments and prove, with a range proof, that a hidden amount is valid without revealing it. Verifying those proofs is expensive, so Elements caches the results: once a proof checks out, its verdict is stored so the node does not have to recompute it. The bug was in how that cache was indexed. According to crypto.news, the cache key was built from the proof bytes and the hidden amount alone, leaving out the asset type and the destination script. Two genuinely different transactions could therefore share a single cache entry.

The attacker exploited exactly that. Over roughly fourteen hours they planted dozens of identical range proofs in ordinary-looking transactions, each costing a few dozen satoshis, so the verdicts landed in the cache. Then they submitted an invalid output that pointed at a poisoned cache entry, and the federation’s nodes skipped the real check and accepted a proof they should have rejected. Chainalysis described the mechanism plainly: the attacker got valid data cached, then submitted invalid data pointing to the same cached result (Chainalysis). The result was almost 4,000 L-BTC conjured out of nothing and pegged out for real Bitcoin.

The timeline is worth reading closely, because it is a textbook example of an N-day exploit, one that uses a bug already fixed in public code but not yet deployed. The patch had been merged into the Elements repository on 2 September; the functionaries were still running version 23.3.3, shipped on 13 April, a gap of nearly five months.

Lightning: The Only Trustless Layer, and It Is Shrinking

The Lightning Network sits at the top of the trust ladder because it is the one design where you never hand your bitcoin to anyone. A Lightning channel is a two-of-two multisig between you and your counterparty, funded by a real on-chain transaction. Payments are just re-signed balances, and if your partner disappears or tries to cheat, you can broadcast your latest state to the base chain and reclaim your coins unilaterally. That is the exit test passed cleanly, enforced by Bitcoin consensus rather than by anyone’s good behavior.

The trade-off is that Lightning is hard to use at scale. It needs inbound liquidity, active channel management, and nodes that stay online, and it works best for streaming small payments rather than storing large balances. Public network capacity peaked around 5,600 BTC in December 2025 and has drifted down toward 4,900 BTC, spread across roughly 41,000 channels and 17,400 nodes, with node counts down sharply from their 2022 peak above 20,000, according to state-of-network data compiled by Spark. We examined that contraction, and the debate over whether it signals decline or consolidation, in our look at whether the Lightning Network is shrinking.

The important point for this discussion is what Lightning is not. It is not a smart-contract platform, and it does not try to be. It moves bitcoin, quickly and trustlessly, and almost everything else in the Bitcoin L2 world is an attempt to add programmability that Lightning deliberately leaves out, which is precisely where the trust starts to creep back in.

Liquid and Rootstock: When a Federation Holds the Keys

Liquid and Rootstock are the two veteran federated sidechains, and both fail the strict exit test by design. On Liquid, launched by Blockstream in 2018, a set of functionaries runs the network and holds the pegged bitcoin under an eleven-of-fifteen multisig; L-BTC is faster and confidential, but you cannot force a withdrawal without the federation’s cooperation. Rootstock, live since 2018 as well, is an EVM-compatible sidechain secured by merge mining, meaning Bitcoin miners validate it alongside the main chain, with RBTC bridged one-to-one through a federated PowPeg. Both traded speed and programmability for a trusted quorum.

For years the federated model looked like a reasonable compromise. Blockstream had never lost customer funds, and the honest-majority assumption held. What the September drain exposed is that a federation protects you against a minority of dishonest signers, but not against a bug that makes every honest signer approve the same fraudulent transaction. The eleven-of-fifteen threshold did its job perfectly and still let $320 million walk, because the flaw lived below the layer the multisig operates on.

Rootstock has stayed quieter and smaller. Its on-chain DeFi holds somewhere around $90 million, concentrated in the veteran protocols Sovryn and Money on Chain, according to DefiLlama. It has never suffered a comparable incident, but it carries the same structural property: the peg is a federation, and the federation is the risk. That is the recurring theme of this whole survey, and it is why the exit test, not the feature list, is the honest way to compare these systems.

Stacks: A Signer Set and the Most Honest Yield in Bitcoin

Stacks takes a different route. Rather than a sidechain, it is a separate blockchain with its own smart-contract language, Clarity, that anchors its history to Bitcoin. Its Nakamoto upgrade in October 2024 cut block times to a few seconds and made Stacks blocks settle with Bitcoin finality, and its sBTC asset bridges bitcoin one-to-one through a decentralized signer set rather than a small federation. That is a genuine improvement on the federated model, but it still does not pass the strict exit test: recovering your bitcoin depends on a threshold of signers behaving, not on a script the base chain will enforce for you alone.

Where Stacks stands out is the honesty of its yield. Its Proof of Transfer mechanism pays people who lock STX in real bitcoin, forwarded from the miners who bid BTC to produce blocks, rather than in freshly printed tokens. In a sector full of emissions dressed up as returns, being paid actual bitcoin for securing the chain is unusual. Stacks co-founder Muneeb Ali has long argued that Bitcoin needs to be made programmable through second layers, so that BTC can be put into smart contracts and into the hands of developers who will build with it (BitcoinWorld).

The numbers have come down with the rest of the sector. sBTC holds around $190 million, well below the $545 million it reached at its first-quarter peak, and the wider Stacks DeFi ecosystem, led by lending markets such as Zest and Granite, sits near $120 million (DefiLlama). Smaller than the hype of 2024 promised, but functioning, and paying a yield that is at least real.

Citrea, BitVM, and the One-of-N Bridge

The most technically ambitious answer to Bitcoin’s scripting limits is BitVM, a scheme first proposed by researcher Robin Linus in 2023 that lets Bitcoin verify arbitrary computation through a challenge-response game rather than by running the computation directly. Its practical payoff is a bridge that needs only one honest participant out of many to stay safe, a meaningful step down the trust ladder from a federation. Citrea, the first production zero-knowledge rollup on Bitcoin, went live on 27 January 2026 and uses a BitVM-based bridge called Clementine to move bitcoin in and out (The Block).

Clementine’s design holds as long as a single signer and a single watchtower remain honest, and in its first six months it processed 150 BTC across the bridge, according to Citrea’s own engineering write-up. The research is moving quickly: a 2026 paper introducing BitVM3, co-authored by Linus and Babylon’s David Tse, uses garbled circuits to cut the worst-case cost of a bridge dispute by around a thousandfold (IACR ePrint 2026/933). The direction of travel is clear, toward bridges that are cheaper to challenge and harder to cheat.

Two caveats keep this from being a finished story. First, Citrea remains tiny, with total value locked in the low single-digit millions of dollars, a research-grade network more than a mass-market one. Second, the most powerful versions of these designs would work far better if Bitcoin adopted a covenant opcode such as OP_CTV or OP_CAT, and that debate is nowhere near settled; no covenant soft fork is likely to activate in 2026. For now, the one-of-n bridge is the best trust-minimized peg the field has, and it is still a work in progress.

Babylon and BTCFi: Staking Is Not a Second Layer

The largest pool of productive bitcoin in 2026 does not sit on any of the layers above. It sits in Babylon, a protocol that lets holders stake native bitcoin to help secure other proof-of-stake chains without wrapping it, bridging it, or moving it off the Bitcoin base chain at all. Babylon leads the Bitcoin DeFi space, which totaled roughly $4.1 billion at the start of September, down about 74 percent from a peak near $9.1 billion in October 2025, according to research by Spark. Earlier reporting put Babylon alone above $4 billion in staked value (The Cryptonomist).

Babylon is important, and it is also, strictly, not a Layer 2. It does not give you a faster way to spend bitcoin; it gives you a way to earn a yield by pledging bitcoin as security, with rewards paid in the BABY token rather than in bitcoin itself. That reward-token design is closer to the staking economics of a proof-of-stake chain than to a payment layer, and it carries the emissions risk that comes with any token-denominated yield. Understanding who actually pays for that security, and whether the reward is sustainable, is the same question we tackled in our breakdown of validator economics.

The broader BTCFi contraction is the real story here. Total value locked fell by roughly three-quarters as liquidity fragmented across isolated pools, as most projects offered copies of Ethereum lending and trading rather than anything new, and as a string of security incidents reminded users that yield on bitcoin still means trusting something. The market has quietly concluded that total value locked without revenue is a vanity metric, and it has repriced accordingly.

Ark and Spark: The Next Off-Chain Generation

The newest wave of Bitcoin L2 design tries to keep Lightning’s self-custody while fixing its usability. Two names lead it. Spark, built by Lightspark, is a statechain-inspired network that is Lightning-compatible and settles Spark-to-Spark payments in under a second; it relies on a set of operators where only one needs to stay honest for users to exit, though at launch only two operators were running it, which is a real centralization caveat. Ark takes a related approach using virtual UTXOs, bundles of pre-signed transactions that share a single on-chain output, where a service provider coordinates rounds but never takes custody, and users can exit unilaterally before their virtual output expires.

What makes this wave interesting commercially is stablecoins. These designs are built to carry assets, not just bitcoin, and Tether led a $5.2 million seed round into Ark Labs in March 2026 explicitly to put USDT on Bitcoin rails, part of a raise that brought the startup’s total backing to about $7.7 million (The Block). Moving dollars over a Bitcoin-secured network is a genuine business, and it is also exactly the activity that sits in regulators’ crosshairs; the compliance stakes are the subject of our piece on how FATF and stablecoins are reshaping the money-laundering rulebook in 2026.

Both designs pass a softer version of the exit test: you can get your bitcoin back without permission, provided you act before a timeout and provided at least one operator stays honest. They are promising, and they are early. As Ark and Spark move from beta toward production, the question is whether they can attract enough operators to be meaningfully decentralized, or whether they simply relocate trust from a large federation to a small handful of well-funded companies.

The Impostors: Wrapped BTC, Botanix, and the eCash Fork

Three things routinely get called Bitcoin Layer 2s that are not. The first is wrapped bitcoin, such as WBTC, which is simply a custodial IOU: a custodian holds your BTC and issues a token on another chain. It is useful and liquid, and it sits at the very bottom of the trust ladder, because if the custodian is hacked, goes insolvent, or is compelled to freeze the assets, your token is backed by nothing you can claim. Wrapped BTC ironically hosts much of the Bitcoin DeFi activity that dedicated L2s were built to capture, which tells you how much most users actually value trustlessness when convenience is on offer.

The second cautionary tale is Botanix, an EVM-style Bitcoin sidechain that shut down in June 2026 about a year after its mainnet launch, despite 100 percent uptime and no security incident. Its team’s post-mortem was unusually candid, concluding, after living inside the project every day, that it simply had not worked, at least not in that market and not on that timeline (CoinDesk). The reason was demand, not technology: most bitcoin holders treat their coins as a reserve asset, and the Bitcoin-DeFi appetite that does exist is already served by wrapped BTC on Ethereum.

The third is the eCash, or ECX, hard fork led by Paul Sztorc’s LayerTwo Labs, which began an opt-in three-stage rollout with an alpha launch on 23 August 2026, a beta scheduled for 20 September, and a permanent launch set for 31 October, timed to the eighteenth anniversary of Bitcoin’s white paper (Bitcoin.com News). It activates Drivechain-style sidechains, but it is a separate chain, not a layer on top of Bitcoin, and at launch it drew about 3.53 petahashes per second of mining power against Bitcoin’s roughly 950 exahashes. A fork inherits Bitcoin’s rules but not its hashpower, which makes it neither more secure than Bitcoin nor a second layer of it.

The Trust Spectrum at a Glance

Pulling the field together, the table below scores the main venues on the properties that matter after the Liquid drain: what kind of system it is, who controls the bitcoin, whether you can exit without permission, and how much value it holds today. The figures move quickly and should be treated as a snapshot, not a live quote.

The pattern is hard to miss. Every system that offers rich programmability sits below the line where you can exit on your own, and every system that keeps you fully in control of your bitcoin does less. That is not a flaw in any one project; it is the shape of the trade-off. The job of a careful user is to know which rung they are standing on before something breaks, not after.

Who Holds Your Bitcoin? The SEC and the Custody Question

For readers in the United States, the Liquid drain lands on top of an unsettled regulatory picture. The Securities and Exchange Commission spent 2026 recalibrating its approach to digital assets, and while spot bitcoin itself is treated as a commodity, the tokens and yield products that Bitcoin L2s generate are a murkier question. A wrapped or bridged asset that promises a return, or a staking product that pays a token, can look a great deal like a security depending on how it is marketed and who controls it. We tracked how the agency is using enforcement and whistleblower incentives in our piece on the SEC’s 2026 enforcement machine.

The deeper issue the hack surfaces is custody. When you hold L-BTC, sBTC, RBTC, or a wrapped token, you are relying on someone else to hold the real bitcoin and honor the peg. That is a custodial relationship in substance, even when the marketing calls it decentralized, and it brings the same questions a regulated custodian would face: who is liable if the reserve is drained, what disclosures are owed to holders, and what happens in an insolvency. Adam Back’s public assurance that Blockstream would cover the Liquid peg was, in effect, the promise a custodian makes, and it was the right one, but it underlined that the backstop was a company’s balance sheet, not a line of Bitcoin script.

None of this makes second layers illegitimate. It means the honest way to describe most of them, for a US audience, is as bitcoin-denominated financial products with counterparties, not as trustless extensions of Bitcoin. The distinction is the difference between a bug being your problem and a bug being someone else’s obligation.

What the Liquid Hack Should Change

The uncomfortable lesson of the Liquid drain is how ordinary its causes were. A fix existed in public for days before the attack; the operators had not deployed it; a caching optimization that most reviewers would skim past hid a mint button. None of that required a nation-state adversary or a novel cryptographic break. It required a patch-management gap and a single overlooked detail in how a cache was keyed. The recovery race that follows any large exploit, chasing funds, negotiating with attackers, and hardening what remains, is a discipline of its own, one we explored in our look at life after the hack.

For anyone using or building on a Bitcoin L2, a few practical rules fall out of the year’s incidents:

• Know your rung on the trust ladder before you deposit, and treat any system without a unilateral exit as a custodial product.
• Watch the gap between when a fix ships and when operators deploy it, because N-day exploits live in that window.
• Prefer systems where the security assumption is small and legible, such as one honest party, over ones where it is a large, opaque committee.
• Discount total value locked as a measure of safety; the Liquid federation held billions and still lost most of a reserve in half an hour.

The good news is that the field is maturing in the right direction. The trust-minimized bridges around BitVM are getting cheaper and stronger, the self-custody wave of Ark and Spark is real, and the market has stopped rewarding empty total-value-locked charts. Bitcoin will keep pushing scale onto other layers, because its base chain is built to stay small. The question every one of those layers has to keep answering, and the one the Liquid drain asked most sharply, is simple: when it breaks, whose problem is your bitcoin?

Frequently Asked Questions

What is a Bitcoin Layer 2?

A Bitcoin Layer 2 is a system that lets you transact in bitcoin without recording every payment on the main chain, while still deriving security from Bitcoin and, ideally, letting you recover your coins to the base chain on your own. Examples range from the trustless Lightning Network to federated sidechains such as Liquid and Rootstock.

Is the Liquid Network safe to use after the $320 million hack?

As of mid-September 2026, Liquid had restarted block production on a patched version of Elements, but peg-ins and peg-outs were still suspended while the reserve was restored. Blockstream chief executive Adam Back said the one-to-one L-BTC peg would be covered. The specific bug that was exploited has been fixed, but withdrawals had not fully reopened.

Which Bitcoin Layer 2 is the most trustless?

By the exit test, the Lightning Network is the most trustless, because you always keep the ability to reclaim your bitcoin on the base chain without anyone’s permission. Newer one-of-n designs such as Spark, Ark, and Citrea’s BitVM bridge are close behind, while federated sidechains like Liquid and Rootstock require you to trust a quorum.

Is wrapped Bitcoin a Layer 2?

No. Wrapped bitcoin is a custodial IOU: a custodian holds your real BTC and issues a token on another chain. It does not derive its security from Bitcoin and offers no unilateral exit, so it sits at the bottom of the trust ladder rather than qualifying as a Layer 2, even though it hosts a large share of Bitcoin DeFi.

Does staking bitcoin on Babylon count as a Layer 2?

Not really. Babylon lets you stake native bitcoin to help secure other proof-of-stake chains without wrapping or bridging it, which is valuable and self-custodial, but it is a shared-security and yield protocol rather than a payment or smart-contract layer. Its rewards are also paid in the BABY token rather than in bitcoin.

By Marcus Okafor, senior markets writer at HOGE Wire.