Infrastructure · Analysis
Osmosis Will Ask Its Voters to Seize an Attacker's 22.65 Bitcoin
A bug in a forwarding mechanism on Nomic — a chain Osmosis does not run and did not write — left roughly 36% of Alloyed BTC without any Bitcoin behind it. The remedy on the table is a governance vote to confiscate what was frozen and spend the treasury on the difference.
The design was supposed to make this impossible. That is the part worth starting with.
Alloyed BTC is Osmosis's answer to a problem every chain with more than one bridge eventually has: users end up holding four incompatible versions of the same asset, liquidity splits four ways, and none of the pools are deep. The alloy takes several Bitcoin representations and presents them as one token, allBTC. Per CryptoAdventure, the four feeding it are native WBTC, Axelar-routed WBTC, cbBTC routed through the Cosmos Hub, and Nomic's nBTC.
Crucially, the alloy is not a blind pool. It enforces what the documentation calls individual composition limits — per-bridge caps that exist for exactly one reason, which is to stop any single bridge's failure from taking the whole token down with it. Diversification is not a side effect of the design. It is the design.
On September 9, Osmosis said that 39.84 nBTC sitting inside the Alloyed BTC contract represented approximately 36% of the alloy's backing, and that those particular vouchers were not backed by Bitcoin at all.
What actually broke, and where
Not Osmosis. This needs saying clearly, because the reflex in a story like this is to blame the chain where the damage showed up.
The flaw was in a custom forwarding mechanism on Nomic. An attacker used it to double-spend nBTC and send false vouchers across to Osmosis over IBC. Osmosis's own contracts behaved correctly. IBC behaved correctly. Both faithfully transported and credited tokens that Nomic had told them were real, which is what they are built to do and the only thing they could have done. The lie was told upstream, and it arrived carrying a valid signature.
The timeline is its own story. According to The Crypto Times, Nomic's last Bitcoin checkpoints were produced on September 6. The chain appeared to halt on September 7. The first public reconstruction of the double-spend transactions came on September 8, from an independent researcher posting as Rarma. Osmosis issued its official statement at 09:48 UTC on September 9.
That is three days between a Bitcoin bridge going quiet and anyone saying why in public — and the first person to say it did not work for either chain.
The freeze, and the arithmetic it leaves
Osmosis's moderation subDAO halted inflows and outflows for both Nomic assets and Alloyed BTC — deposits, withdrawals, minting and redemptions, per PANews. Validators then coordinated an emergency chain upgrade that quarantined 22.65 BTC sitting at an address linked to the attacker before it could be moved.
The Osmosis chain itself was never halted, and existing liquidity pools stayed tradable. Only the alloy's mint-and-redeem machinery stopped. That distinction matters: nobody was locked out of the chain, but anyone holding allBTC was, for the duration, holding a claim that could not be exercised.
The recapitalisation gap, in BTC
| Line | Amount | Status |
|---|---|---|
| Unbacked nBTC in the alloy | 39.84 | ≈36% of allBTC's backing |
| Frozen at attacker address | 22.65 | Locked by emergency upgrade; seizure needs a vote |
| Remaining shortfall | ≈17.19 | Proposed source: the community pool |
So the plan has two halves, and only one of them is technical. Recovering the frozen 22.65 BTC closes most of the hole. The remaining 17.19 or so has to come from somewhere, and the somewhere Osmosis has named is its own treasury. Sunny Aggarwal, a former Osmosis co-founder, said the community pool holds enough Bitcoin to cover the shortfall if governance approves the seizure route.
A vote to confiscate
Strip the vocabulary away and the proposal is this: OSMO holders will be asked to authorise taking Bitcoin out of an address the chain does not control the keys to, on the grounds that the chain's validators believe it belongs to an attacker.
There is a defensible case for it. The funds are demonstrably the product of vouchers that were never backed; leaving them alone means honest allBTC holders absorb the loss while the attacker keeps the proceeds; and the freeze has already happened, so the vote ratifies a decision the validators made under time pressure rather than initiating one. Chains have done versions of this before, and the alternative — socialising the whole 39.84 onto the treasury while the attacker walks — is worse on every axis except procedural purity.
It is still worth being precise about what it is. A network of token holders voting to move someone else's balance is a capability, and capabilities do not stay pointed at the people you first aimed them at. The narrow question in front of Osmosis governance is whether to make allBTC whole. The wider one, which no proposal text will contain, is what the precedent is worth the next time the address in question is merely suspected.
At the time the outlets covering this filed, no proposal ID had appeared on public governance dashboards. The seizure is announced, not yet drafted.
What has not been said
Nomic has not published a public account of the failure. There is no post-mortem, no statement of when the forwarding bug entered the code, and no explanation of the gap between the last checkpoint on September 6 and the first outside analysis on September 8. Osmosis has promised a post-mortem; it had not appeared as of this writing.
One thing has been clarified. An Osmosis Foundation conversion of allBTC to WBTC drew immediate suspicion of an insider exit; Aggarwal said it was tied to a scheduled BitGlobal test of native WBTC burns rather than an opportunistic move. That is his account, offered on X, and the transactions are on-chain for anyone who wants to check the schedule against them.
The Take
The composition limits worked exactly as advertised, and that is the uncomfortable finding. The cap did its job — Nomic's failure took roughly 36% of the alloy, not 100% — and 36% is still catastrophic for anything calling itself Bitcoin-backed. A design that converts a total loss into a one-third loss is a real improvement over the alternative, and it is nowhere near enough when the product's entire promise is one-to-one redemption. Alloyed assets solve a genuine problem, and they solve it by making the weakest bridge in the set everyone's problem instead of only its own users'. Anyone who chose native WBTC on purpose, specifically to avoid bridge risk, still woke up short on September 9, because the alloy had spent their exposure on their behalf. That is not a bug in the implementation; it is what pooling means. The honest fix is not a better cap. It is telling holders, in the interface and not in the docs, which bridges they are underwriting today and at what weight — and letting the people who wanted the boring wrapper keep holding the boring wrapper.