AI Cuts Crypto Attack Costs by 99.9%, Forcing Bitcoin Quantum Defense Overhaul

Woofun AI reports that Anthropic’s deployment of multi-agent AI systems has fundamentally altered the cost structure of cryptanalysis, identifying critical vulnerabilities in the HAWK post-quantum signature scheme and enhancing attacks on AES, with each breakthrough costing $100,000 in compute. These findings do not currently compromise Bitcoin, Ethereum, hardware wallets, or production encryption standards, as HAWK remains undeployed and the AES work targets a reduced seven-round version of a ten-round cipher.

However, the ability to compress complex cryptanalytic processes into days using specialist mathematical software and API compute signals a paradigm shift in how emerging cryptography must be tested, particularly as Bitcoin researchers navigate a multi-year transition away from signatures vulnerable to quantum computers. Any replacement algorithm will now face automated analysis alongside conventional human review, raising the bar for security validation in NIST’s ongoing standardization efforts.

The HAWK experiment utilized a multi-agent system named Claude Mythos, which operated semi-autonomously to develop and test an attack over 60 hours. The agents were capable of searching academic literature, writing Python and Sage code, running mathematical experiments, and sharing promising results across the network. In one notable instance, an approach initially rejected by one agent was pursued by another until it yielded the decisive breakthrough. Human direction was limited to organizing the work, selecting tools, and verifying the resulting mathematics, with the operator possessing a theoretical computer science background but no specialism in lattice cryptography. The entire experiment consumed $100,000 in API compute, a substantial expense that remains within reach of well-funded research laboratories, security companies, and sophisticated adversaries, thereby democratizing access to high-level cryptanalytic capabilities.

HAWK is a lattice-based scheme participating in the third round of NIST’s additional post-quantum signature process, a category of algorithms built around mathematical problems that known quantum methods are not expected to solve efficiently. Before Anthropic’s intervention, HAWK had undergone two years of public examination without such a discovery. According to Anthropic’s research, Claude Mythos identified an overlooked symmetry in HAWK’s underlying lattice structure, exploiting it to create a more efficient route to recovering the secret key.

For HAWK-256, a deliberately small challenge parameter designed for cryptanalysis, the estimated attack cost fell from 2^64 operations to 2^38. While the larger proposed versions remain impractical to break and the weakness does not apply to every lattice-based design or other candidates in NIST’s process, the finding indicates that HAWK requires substantially stronger parameters to recover its intended security level.

Anthropic estimates that to maintain security, HAWK’s keys may need to double in size, which would significantly impact blockchain efficiency. HAWK-512 already utilizes a 1,024-byte public key and a 555-byte signature; increasing these figures would add storage, bandwidth, and verification costs, eroding the efficiency advantage that made the scheme attractive. On a blockchain, such overhead multiplies across transactions and is permanently replicated by nodes, meaning a mathematically sound upgrade can become impractical if its signatures consume too much block space. This trade-off between security margin and operational efficiency is a critical variable for any post-quantum candidate, as the network must balance robustness against the finite resources of block space and node storage.

Bitcoin Improvement Proposal 361, or BIP-361, outlines how the network could migrate away from the ECDSA and Schnorr signatures that currently protect transactions. The proposed sequence involves restricting legacy coins to post-quantum destinations, followed later by tighter recovery rules for funds that have not moved. The transition could span five years after activation, though BIP-361 remains a draft with no activation date and leaves the replacement scheme to a separate proposal. Choosing that algorithm involves more than confirming resistance to known quantum attacks; developers must weigh signature size, verification speed, hardware wallet support, block-space demand, and the possibility that later research forces stronger parameters, as demonstrated by the HAWK case.

Woofun AI data shows that in its second project, Anthropic targeted AES-128, which protects data through ten rounds of repeated mathematical transformations. Claude worked against a seven-round research version under assumptions that require 2^105 chosen plaintexts, conditions that render the attack unusable against real encrypted data. Despite these constraints, the system improved the previous best method by 200 to 800 times, depending on how the computational cost is measured. Its main contribution was a fingerprinting technique called the "Möbius Bridge," which removed a 256-way guess required by earlier approaches. This improvement was achieved through an autonomous framework in which multiple agents proposed hypotheses, ran experiments, and discarded failures, showcasing the ability of AI to contribute original ideas to mature areas of mathematical cryptanalysis.

Reaching the AES result required an autonomous framework that generated around one billion output tokens and consumed $100,000 in API usage, with the central insight produced within a week. Two human researchers then spent close to a month checking the mathematics and gaining confidence that the attack was valid, highlighting a significant validation gap. Automated systems can pursue many speculative ideas at once, while qualified cryptographers must reproduce each credible result, determine whether it is genuinely new, and assess its practical consequences. That gap between production and validation is where security teams will feel the strain, as a growing volume of machine-generated findings will include mistakes, rediscoveries of known work, and a smaller number of serious vulnerabilities, requiring as much investment to sort as to produce.

The strategic risk lies in treating a post-quantum label as a permanent guarantee, as a design may remain unbreakable in practice while losing enough of its security margin to become inefficient or unsuitable for deployment. Bitcoin’s eventual upgrade should therefore preserve room for another change, as building wallets and protocol rules around one supposedly permanent replacement could leave the network facing a second disruptive migration if that scheme is later weakened. Crypto projects will need clear submission channels, independent reproduction, and strict severity standards, ensuring that an unverified model output does not trigger a protocol change, though dismissing every machine-generated report risks overlooking a genuine flaw. This dynamic affects zero-knowledge proofs, wallet recovery methods, cross-chain protocols, and decentralized networks, which may need months to coordinate an upgrade once a weakness becomes public.

Future requirements for Bitcoin’s quantum defense include parallel testing of several candidates rather than building the entire migration around an early favorite. Evaluations must cover mathematical security, transaction weight, verification costs, and compatibility with exchanges, custodians, and hardware devices. AI-assisted research can form part of this process alongside public competitions, academic review, and implementation audits, with human experts confirming every finding before it influences consensus rules. Software architecture also matters, as replaceable cryptographic modules would give wallets and nodes a safer route to change algorithms if new research weakens the first standard adopted, mitigating the coordination problem among developers, miners, node operators, and users holding coins that may not move for years.

Anthropic has not shown that AI can break Bitcoin’s existing signatures, nor does Claude perform the work of a quantum computer; its impact lands earlier in the process, challenging proposed algorithms and exposing hidden mathematical structure before deployment. Quantum computing remains the eventual threat to operational wallets, while AI reshapes the research environment in which Bitcoin’s defense must be designed and tested. The HAWK result therefore says more about the standard Bitcoin’s future cryptography must meet than about any single candidate, requiring a replacement to withstand quantum theory, conventional cryptanalysis, and automated research systems capable of exploring ideas faster than human reviewers can follow.