Blockstream's Liquid Network held more than 4,200 bitcoin in its peg wallet on Sunday, September 6. By that evening it held about 197. The roughly 4,000 BTC that left, worth close to $320 million at the time, was withdrawn with valid signatures from eleven separate institutions, every one of which was doing exactly what it was designed to do.
That last detail is why this incident deserves more attention than it got. 2026 has been the year of the stolen key, with compromised credentials overtaking smart contract bugs as the leading cause of DeFi losses for the first time on record. Liquid lost its reserves with the keys working perfectly.
How the Reserves Walked Out
Where the phantom coins came from
Liquid is a Bitcoin sidechain. Users lock BTC into a federated peg wallet and receive L-BTC on the sidechain at one to one, and the whole arrangement rests on a single invariant: the amount of L-BTC in circulation should never exceed the amount of BTC held in the peg.
An inflation bug in the sidechain's consensus code broke that invariant. Attackers exploited it to create over 4,000 L-BTC that had never been backed by anything, then submitted an ordinary peg-out request to redeem those tokens for real bitcoin. On paper the request was indistinguishable from a legitimate one, because by the rules the network was actually running at that moment, the tokens were legitimate.
Why eleven signers agreed
The peg wallet is protected by an 11-of-15 multisig held by vetted federation members, exchanges and institutions known as functionaries, each running the signing logic inside a hardware security module. Eleven of them signed. Their HSMs checked what HSMs check: that the peg-out was well formed and that the sidechain's consensus rules considered the L-BTC being burned to be valid. The consensus rules were the compromised component, so the verification passed cleanly.
An 11-of-15 multisig answers one question: did enough authorized parties approve this withdrawal. Whether the withdrawal was backed by anything sits outside its remit entirely.
This is worth dwelling on, because the reflexive fix after every custody failure is to add signers. Fifteen reputable institutions with hardware-isolated keys is close to the best version of that answer available, and it bought nothing here. Distributing signature authority protects against a corrupt or compromised signer. It offers no defense at all when the bug lives upstream of the signature.
The Bounty That Set Its Own Price
The attackers left an on-chain message declaring themselves white hats, and after Blockstream confirmed a patch they returned around 3,400 BTC, keeping roughly 598 BTC, about $47 million, as a self-appointed bounty.
Fifteen percent, unilaterally assessed, non-negotiable, and paid in advance. Security researchers who go through a disclosure process and wait six weeks for a $50,000 payout are entitled to feel some professional envy. Ledger's CTO and a good portion of the security community declined to accept the framing, which is the correct response to a fee schedule you had no part in agreeing to. The practical reality is that Blockstream recovered 85% of a total loss, and the alternative was worse, which is precisely the leverage the whole performance depends on.
Every Peg Has the Same Unchecked Invariant
Here is the part that generalizes beyond one sidechain. The supply-versus-reserves invariant is the single most important number in any wrapped or pegged asset, and almost nowhere is it checked continuously by anyone with the ability to halt withdrawals.
Reserve attestations exist, and they are published monthly or quarterly, snapshotted at a date convenient to the issuer. An inflation bug drains a peg in one block. The technical post-mortem traced the mint and the peg-out through the same afternoon, which means a quarterly attestation would have documented the theft roughly eleven weeks after the money was gone.
What would have caught it is embarrassingly simple: a monitor comparing circulating L-BTC against peg-wallet BTC every block, wired to a circuit breaker that freezes peg-outs on any divergence. That check costs almost nothing to run. Its absence across most of the wrapped-asset market is a design choice that nobody made deliberately, which is usually how these things go.
The LeveX Take
Treat every wrapped or pegged asset as carrying two separate risks that get quoted as one. There is the custody risk, meaning whether someone can steal the collateral, and there is the accounting risk, meaning whether the token supply still corresponds to the collateral. The industry has spent a decade hardening the first and almost no effort instrumenting the second, and Liquid is the clearest demonstration yet that the second can bypass the first entirely.
The practical consequence for traders is a pricing question that most people never ask. A wrapped asset trading at parity with its underlying is quoting zero probability of an accounting failure. That is a real position with a real payoff profile, and holders of these instruments are short that risk without being paid for it. The moment a supply divergence becomes visible, the wrapper gaps rather than drifting toward a discount, because peg-outs get suspended and the exit closes before the price finishes moving. Anyone weighing a yield spread on a wrapped instrument against native custody should read our comparison of wrapped Bitcoin versus native Bitcoin with that gap risk in mind.
This is the argument for Proof of Reserves as a live instrument rather than a marketing badge. LeveX publishes 1:1 Merkle-tree reserve verification that lets any user check their own balance is included in the attested total, currently at 111% for BTC, 149% for ETH, and 160% for USDT. Over-collateralization is the part that matters for a failure like this one, because a peg running at exactly 100% has no absorption capacity when the accounting turns out to be wrong. The question to put to any venue or wrapper holding your coins is who is counting the tokens, and how often. Signer counts are the easy answer, and the one the industry already solved.
What to Ask Before You Hold a Wrapper
Bitcoin's Layer 2 and sidechain ecosystem has spent years selling the idea that you can have Bitcoin's security guarantees plus somebody else's feature set. September 6 priced that claim properly. Liquid's underlying bitcoin was never at risk from a Bitcoin protocol failure, and it left anyway, because the security you actually inherit is the security of the weakest consensus layer between you and your coins.
The follow-through to watch is whether Blockstream publishes a continuous supply-monitoring commitment alongside its patch, and whether the other federated pegs and bridge operators quietly ship the same check in the next month. Given that 2026 already set a record number of crypto hacks, the honest expectation is that most of them will wait for their own turn.
For traders who would rather hold the asset itself than a claim on it, LeveX runs BTC spot and futures markets with reserves published and independently checkable, and the Crypto in a Minute series covers how pegs, wrappers, and bridges differ under the hood.