TL;DR:
- The Claude Mythos Preview model from Anthropic identified an attack that accelerates the breaking of reduced seven-round versions of the AES algorithm by 200 to 1,000 times.
- The artificial intelligence tool halved the key strength in HAWK, a candidate signature scheme for NIST’s post-quantum standardization.
- The Bitcoin network currently uses the ECDSA elliptic curve digital signature algorithm to validate and authorize transactions on the blockchain.
Recently, Anthropic announced that its Claude Mythos Preview model discovered new mathematical vulnerabilities in cryptographic algorithms. The Anthropic research demonstrates that advanced language models could evaluate the mathematical security of encryption schemes before quantum computers manage to compromise current signatures on networks like Bitcoin.
Advances in cryptanalysis and the impact on digital signatures
The official report presented by Anthropic details that the Claude Mythos Preview model developed an attack method against a simplified seven-round version of the Advanced Encryption Standard (AES-128). According to measurements provided by researchers, the mathematical technique discovered by the AI system increased the attack speed by 200 to 1,000 times compared to theoretical procedures known to date.


The published results suggest that frontier models are acquiring capabilities to generate autonomous and unprecedented cryptographic research. According to documentation provided by Anthropic, the same model structured an improved attack against HAWK, a digital signature scheme specifically designed to operate in a quantum-resistant environment.
The test executed on the HAWK algorithm required approximately 60 hours of processing and represented a cost of nearly $100,000 in API consumption. According to data released by the tech firm, the operation managed to halve the effective key strength in the HAWK-256 test parameter by finding symmetries in the lattice structure.
In the current architecture of the Bitcoin network, the protocol employs the elliptic curve digital signature algorithm, known as ECDSA, to authenticate ownership of funds and validate blocks. According to various computer science research reports, the emergence of quantum processors with a sufficient number of physical qubits could compromise the security of ECDSA in the long term.
For this reason, Bitcoin core developers initiated technical debates on the gradual transition toward signature schemes resistant to quantum attacks. According to industry analysis documents, algorithms based on lattice problems are positioning themselves as the most promising alternatives to replace ECDSA in future network upgrades.
Despite the attention generated by these findings, the Anthropic research team stressed that current computer systems do not face an immediate threat. The computational attacks were applied exclusively to weakened schemes for testing purposes and not to code deployed in production environments.
Similarly, company documentation confirms that the full ten-round version of the AES cipher has not been compromised and remains secure for global data traffic. In this regard, Peter Van Valkenburgh, Executive Director of Coin Center, stated that these advances present a scenario where developers must consider both the progress of quantum hardware and the advancement of artificial intelligence algorithms when evaluating the security of new standards.
The United States National Institute of Standards and Technology (NIST) will keep the evaluation process for post-quantum candidates open during its upcoming technical conferences scheduled for the end of the year. In parallel, cryptography research groups plan to conduct additional tests using automated tools to verify the robustness of proposed algorithms before their eventual adoption in blockchain networks.





Be the first to comment