Cut Through CryptoBetaSubscribe
Bitcoin quantum threat escalatesLatest of 12
Claude Guts A Post-Quantum Bitcoin Candidate

Claude Guts A Post-Quantum Bitcoin Candidate

BitcoinJul 29, 2026

Institutions, developers, and regulators are racing to assess and implement post-quantum cryptographic protections for Bitcoin and broader crypto infrastructure.

Anthropic announced that Claude, working largely autonomously, took roughly 60 hours to find a weakness in HAWK — a proposed "quantum-proof" digital signature design — that two years of expert review had missed, making its smallest version about 67 million times cheaper to crack. No systems in use today are affected, but crypto media framed the finding as a warning for Bitcoin's long-term quantum defences.

Quantum Bitcoin: the audit just got faster than the auditors

Some background in thirty seconds. Every Bitcoin transaction is approved with a digital signature — a piece of math that proves you own your coins without revealing the secret key that controls them. Quantum computers, if they ever get big enough, could break that math. So the US government's standards body (NIST) has been running a years-long public contest to find "quantum-proof" replacements, with cryptographers worldwide submitting designs and trying to break each other's work. HAWK was one of the contestants — attractive because its signatures were unusually small, which matters on a blockchain where every byte costs money.

Last week, Anthropic pointed its AI at HAWK. In 60 hours and about $100,000 of computing, Claude found a hidden shortcut that human experts had missed for two years — making the smallest version of HAWK roughly 67 million times cheaper to crack. The scheme isn't dead: larger versions still hold up, and the fix is straightforward — use bigger keys. But bigger keys erase the small-signature advantage that was HAWK's entire reason to exist.

Cue the "Bitcoin's quantum shield just broke" headlines. Except Bitcoin was never going to use HAWK. The current draft plan for going quantum-proof (a proposal called BIP-360) would let coins migrate to new address types built on the schemes NIST has already finished testing and approved — and Claude's attack doesn't touch those. Nothing anyone actually uses got broken. If anything, Bitcoin's habit of choosing the boring, battle-tested option just aged well.

What did change is what "vetted by experts" is worth. Two years of the world's best cryptographers reviewing HAWK missed what an AI found in three days. That cuts both ways. It's now harder to trust any quantum-proof scheme just because humans reviewed it — but it's also possible, for the first time, to have machines attack these designs at a speed and scale humans never could, before anything is locked in. For Bitcoin, which moves slowly by design and realistically gets one shot at this migration, that's a trade worth taking: less faith in credentials, better tools for earning it.

Related storylines

7 sources