What a quantum attack on bitcoin would actually look like

Most descriptions of a quantum attack on bitcoin are either a headline or a formula. Neither tells you what would happen, in order, on a chain you can inspect yourself. This article walks through it step by step: what an attacker would need, what they would do, what the rest of us would see, and what would stay exactly as it is. It uses the terms and the framing of the current draft proposals, it gives no dates, and it says plainly where the answer is that nobody knows.

Gerard Sala16 November 20269 min read

The short version. A quantum attack on bitcoin would not break the network, the hashes or the rules. It would produce an ordinary, valid transaction, signed with a private key that someone other than the owner worked out from a public key. There are two forms. In a long exposure attack the public key is already sitting on the chain, so the attacker has as much time as they want. In a short exposure attack the key appears only when the owner spends, and the attacker has to derive it and get a competing transaction mined before the original confirms. The first needs a machine that works; the second needs one that is also fast. Nodes could not tell either from a normal spend.

What breaks, and what does not

Bitcoin signatures use the secp256k1 elliptic curve. A private key is a number, and the public key is a point derived from it. Going from the point back to the number is the discrete logarithm problem, which classical computers cannot solve at this size. Shor's algorithm can, on a large error-corrected quantum computer. BIP-360, the draft that proposes Pay-to-Merkle-Root, calls this the primary threat and gives it a name: quantum key recovery, the derivation of private keys from public keys.

That is the whole attack surface. Hashes are a different kind of problem, and we covered why they hold up in Grover vs Shor. So everything below starts from one input: a public key the attacker can read.

How big a machine? The most cited gate-level estimate, Roetteler, Naehrig, Svore and Lauter (2017), bounds the job for a 256-bit prime field at 2,330 logical qubits and about 1.3 × 1011 Toffoli gates. Those are error-corrected qubits, each built from many physical ones, and no such machine exists today. What matters here is that it is a computation of fixed size, run once per key. Each public key is its own job.

The long exposure attack, step by step

BIP-360 defines a long exposure attack as one on keys that stay exposed for longer than it takes to confirm a spend, and says these are likely to be the earliest quantum attacks made possible on Bitcoin, because the attacker has as much time as the key remains exposed.

Step 1: choose a target. No intrusion is needed. The chain is public, and the list of exposed keys can be built by anyone with a node. BIP-360 lists the output types it calls fundamentally vulnerable to long exposure attacks:

For scale, BIP-361 states that over 34% of all bitcoin had a revealed public key as of 1 March 2026. Our own report on the 100 largest addresses, a snapshot from 3 September 2026, found 67 of them with their key on the chain.

Step 2: compute the private key. This happens entirely off the network. The attacker copies a public key and runs the computation on their own hardware. Nothing is broadcast, no node is contacted, and nothing on the chain changes. How long one key would take on a real machine is not known, and for this attack it hardly matters: the key is not going anywhere.

Step 3: sign and broadcast. With the private key in hand, the attacker builds a normal transaction that sends the coins to an address of their own, signs it, and broadcasts it. The signature is valid, because it was made with the right key. Every node accepts it, a miner includes it, and it confirms like any other payment. No rule is broken and no block is rewritten. To the protocol, the owner spent their coins.

What it would look like on the chain

This is the part that is easiest to get wrong. The visible event would be a spend from an address whose key was already public. That is all. A signature does not record who made it or how the key was obtained.

So the first sign could be ambiguous. Old coins move for ordinary reasons. If a long-dormant exposed address spent tomorrow, observers could not prove from the chain alone whether the owner or someone else signed. That is our reading, and BIP-361 goes further. Its authors write that attackers could compute keys for known public keys and move the funds weeks or months later, in what they call a covert bleed, so that the moment a capable machine exists might only be known much later. They also note that it is impossible to know an attacker's motives in advance: one who wants money would try to stay unnoticed, and one who wants damage would not.

Two more things follow from the mechanics. Because each key is a separate computation, an attack would arrive as a series of individual spends, not as every exposed coin moving in one block. And because the cost per key does not depend on what the key holds, it is reasonable to expect large exposed balances to be more attractive than small ones. That second point is an inference of ours, not a finding from any source.

A confirmed theft would also be as final as any other transaction. Bitcoin has no mechanism to reverse a valid spend.

The short exposure attack: a race in the mempool

Coins on a hashed address that has never spent show only a hash, and Shor's algorithm has nothing to start from. But to spend them, the owner has to reveal the public key inside the transaction. From the moment that transaction is broadcast until it is confirmed, the key is visible to every node that relays it. BIP-360 calls an attack in that window a short exposure attack.

Step 1: watch the mempool. Unconfirmed transactions are public. The attacker picks one whose input reveals a key guarding a worthwhile amount.

Step 2: derive the key before the transaction confirms. Blocks arrive about every ten minutes on average, but the gap varies, and a transaction that pays a low fee can wait much longer. The attacker's deadline is whenever the original gets mined.

Step 3: broadcast a conflicting transaction. It spends the same coins to the attacker's address and pays a higher fee. Since Bitcoin Core 28.0, nodes accept such replacements by default: the release notes record the mempoolfullrbf option changing from 0 to 1. Miners have a direct incentive to pick the version that pays more.

Step 4: one of the two confirms. If it is the attacker's, the owner's payment never happens. If it is the owner's, the coins have already moved and the revealed key guards nothing, provided the address is not used again.

This is a much harder attack. BIP-360 says it requires faster machines than a long exposure attack and considers it unlikely that early ones will be fast enough. The same text is equally clear about the other side: these attacks cannot be prevented through wallet hygiene, because revealing a public key is necessary for spending. A full fix needs post-quantum signatures, which BIP-360 leaves to a separate, future proposal.

Some setups stretch the window from minutes to days. A Lightning channel that closes with a timelock leaves a key on the chain while the delay runs, which we looked at in is Lightning quantum safe.

What it would not look like

  • Not every wallet at once. Coins on hashed addresses that have never spent are not reachable by a long exposure attack. Only keys that are visible can be worked on, and one at a time.
  • Not a broken blockchain. Blocks, proof of work and transaction history rest on hashes. Nothing here rewrites the ledger or lets anyone create coins.
  • Not something a node could filter. There is no marker that separates a signature made by the owner from one made with a derived key.
  • Not undone by going quiet. A key that is on the chain stays there. The coins behind it stop being exposed only when they move to an address whose key has never been shown.

What nobody can tell you

Whether a machine of the required size will be built, and when, is an open question, and we do not put a date on it. How long a single key recovery would take on such a machine is also unknown, and that number is what decides whether a short exposure attack is practical at all. Nobody can say whether the first capable operator would act openly, quietly, or not against bitcoin at all. And nobody knows what the protocol will look like by then: BIP-360 and BIP-361 are drafts, neither is activated, and both may change. We read the current text of each before writing this, and summarise them in what is BIP-360 and what is BIP-361.

What this means for your coins

The step-by-step view makes the practical advice short. Step 1 of the long exposure attack needs a visible key, so the useful question is whether yours is visible.

  • Check which of your addresses have spent. Any balance left on one of them sits on a public key.
  • Move those coins to a fresh address that has never spent, and do not reuse it. The move itself reveals the old key for the time it waits to confirm. That is short exposure, the harder attack, and it ends with the coins behind a hash.
  • Know what Taproot means. A bc1p address shows a key as soon as it receives.
  • Follow the proposals, not the headlines. Protection against short exposure depends on protocol changes that do not exist yet.

Monitoring has a narrow role here, and it is worth being exact about it. An alert on an address can tell you that it has spent or that its balance has moved, after the transaction is visible. It cannot see anyone computing a key, because that leaves no trace, and it cannot stop or reverse a transaction. It detects; it does not block. Our first article covers the wider picture.

See whether your key is already visible

Paste an address, or an extended public key for a whole wallet, and ShorWatch checks whether each address shows only a hash or has its public key on the chain, and shows the balance on each side. It runs in your browser against public chain data from mempool.space. Nothing is sent to us and nothing is stored.

Scan an address

Read-only · Non-custodial · Nothing stored · No account needed

Limits. The scanner reports exposure at rest from public chain data: whether a public key is already visible on the chain. It says nothing about short exposure, which applies to every spend, and it cannot tell whether a key leaked off the chain, for example through a shared extended public key or Lightning gossip. It does not cover pre-2011 pay-to-public-key outputs; it treats any multisig script address as opaque until it spends; and it cannot see coins held on an exchange or a custodial wallet in someone else's addresses. Alerts detect changes after they appear on the chain; they do not block anything. This is an informed first pass, not an audit, and not financial or security advice. The qubit and gate figures are upper bounds from Roetteler et al. (2017) and have been revised by later work. Figures from the top-100 report are a snapshot from 3 September 2026; the BIP-360 and BIP-361 statements are taken from the drafts as published on 16 November 2026.