Liquid Network’s federation wallet lost 3,998.5 BTC on Sept. 6, leaving 207.275 BTC from a balance of about 4,200 BTC before the incident, according to BlockTempo. That means roughly 95% of the wallet was emptied. The attacker later returned 3,400 BTC, but still held 598.5 BTC, which the report valued at about $47.61 million at current prices.
Much of the recent discussion has focused on whether the party that returned most of the funds should be treated as a white-hat actor. BlockTempo pointed to another detail: none of the nearly 4,000 BTC left the Bitcoin mainnet. Bitcoin blocks continued to be produced as usual, consensus was not altered, and the problem sat in the mechanism that held and redeemed the BTC.
The report also said Liquid’s defenses were not defeated through key theft. The attacker did not steal any private key.
All three peg-out checks passed
Liquid, built by Blockstream, is a Bitcoin sidechain. Users lock BTC into the federation wallet and receive L-BTC on the sidechain. To move back to mainnet BTC, assets go through the peg-out process, which normally involves three checks:
- L-BTC on the sidechain is burned first;
- at least 11 of the federation’s 15 members sign using their own hardware security modules;
- the transfer then passes PAK whitelist authorization.
In this case, all three steps were completed normally. The flaw, BlockTempo said, was in Elements, the Bitcoin Core fork used by Liquid nodes. Its cached rangeproof validation indexed data using only the proof itself and the value commitment. That let the attacker create L-BTC without matching BTC backing.
Those tokens were then used in peg-outs. Liquid’s automated systems could not distinguish the forged assets from valid ones, so the signing flow continued and real BTC was paid out.
BlockTempo noted that the 11-of-15 threshold was designed to guard against malicious or unavailable federation members. Official documentation says it would take at least five members going offline at the same time to disable the multisig setup. In this incident, not a single member failed. The threshold controls who is allowed to sign, not whether a given transaction should have been signed in the first place.
Liquid’s scale was far below the largest wrapped-Bitcoin products
Before the incident, Liquid’s roughly 4,200 BTC equated to about $334 million at current prices, the report said. In the broader market for BTC wrappers and related products, that put Liquid in the lower tier.
Using Sept. 9 data from DefiLlama and CoinGecko, BlockTempo listed the following products:
- WBTC: $9.23 billion, or 116,134 BTC, custodied by BitGo;
- cbBTC: $7.87 billion, or 98,823 BTC, issued by Coinbase;
- BTCB: $5.41 billion;
- Lombard’s LBTC: $687 million;
- Function’s FBTC: $682 million;
- Kraken: $645 million;
- SolvBTC: $509 million;
- Threshold’s tBTC: $354 million;
- Stacks’ sBTC: $193 million.
WBTC alone was 27 times the size of Liquid. BlockTempo said the reason this incident stayed around the $300 million range had little to do with security design and more to do with the fact that Liquid was not especially large to begin with.
Three trust models for wrapped BTC
The report grouped these products into three trust assumptions.
The first is the custodial model, represented by WBTC, cbBTC and BTCB. Users trust a company to hold the underlying BTC, so failure tends to take the form of company-level risk.
The second is the federation model, represented by L-BTC. Here, users trust a group of companies with known identities not to collude. This incident showed that collusion is not the only way the model can fail.
The third is the threshold-signature model, with tBTC cited as an example. In that setup, randomly selected signers jointly manage private keys, post collateral and face slashing if they act maliciously. Trust rests on code and economic incentives.
Each extra layer adds another failure assumption
BlockTempo also framed a common question in practical terms: how many custodial layers sit between a BTC on the mainnet and the place where that BTC eventually earns yield?
If a user moves self-custodied BTC into WBTC and then uses it in lending, that is one layer of trust, centered on BitGo. If the same WBTC is bridged to another chain and deposited into a protocol, the stack rises to three layers because the bridge and the destination protocol each add another assumption.
Restaking-style products such as Lombard’s LBTC and SolvBTC sit on top of wrapped assets to begin with, so the layering goes higher.
Each additional layer introduces another set of assumptions that can fail, and those assumptions do not cancel one another out. In Liquid’s case, the failure was not in multisig, not in custody, and not in private-key storage. It was in the indexing logic for validation cache data.
Network still paused
As of publication, the Liquid network was still paused, bridge nodes were disabled, and exchange deposits and withdrawals for L-BTC had not resumed. Blockstream had not said who would ultimately absorb the 598.5 BTC shortfall.
The report closed on one final point: the roughly 4,000 BTC remained on the Bitcoin mainnet throughout the episode. What changed was the custody path around them, which briefly shifted hands and then moved back.
FAQ
How much BTC was taken from Liquid Network, and was any of it returned?
On Sept. 6, 3,998.5 BTC was taken from the federation wallet, worth about $318 million according to the report, leaving 207.275 BTC. The attacker later returned 3,400 BTC but still kept 598.5 BTC, valued at about $47.61 million at current prices.
What trust risks come with wrapped BTC?
BlockTempo divided wrapped-BTC designs into custodial, federation and threshold-signature models. It noted that WBTC’s $9.23 billion is held by BitGo, while cbBTC’s $7.87 billion is held by Coinbase. Liquid showed that assets can still be moved even when no multisig key is stolen.