AI Agents and Researchers Cut Quantum Attack Resource Cost by 86%

Researchers and AI coding agents have reduced the resource benchmark for a critical component of a quantum attack on Bitcoin and Ethereum by 86%. The effort, part of the ECDSA.Fail challenge launched by Eigen Labs, focused on optimizing the quantum circuits required for elliptic curve point addition—a vital step in executing Shor's algorithm.

By July 26, the resource score for the quantum circuit dropped from 10.75 billion to 1.496 billion. This benchmark, which represents the product of qubits and Toffoli gates, is approximately half of a previous estimate set by Google Quantum AI, though researchers note that direct comparisons are complicated by differing accounting methods. The challenge spanned roughly eight weeks and involved over 100 participants, including human researchers and AI agents.

The Path to Q-Day

Bitcoin and Ethereum both rely on the secp256k1 elliptic curve (ECDSA) to secure transaction signatures. While these systems are currently secure against classical computers, a sufficiently powerful fault-tolerant quantum computer could use Shor's algorithm to derive private keys from public keys. This theoretical event, often called "Q-Day," would allow an attacker to steal funds from any address where the public key is known.

To mitigate this existential threat, the National Institute of Standards and Technology (NIST) is currently standardizing post-quantum replacements, and various firms are investing in quantum-proof cryptographic research to ensure blockchain longevity.

Implications for Blockchain Security

The dramatic reduction in estimated resources suggests that the theoretical "cost" of a quantum attack is decreasing more rapidly than the physical hardware is advancing. This trend potentially shortens the window available for major blockchains to migrate their entire user bases to post-quantum cryptography before a viable attack becomes possible.

Beyond the security implications, the project serves as a proof of concept for "Open Autoresearch." According to the ECDSA.Fail researchers, the project provided a case study in a verifier-gated research process where humans and AI agents iteratively generated, tested, and shared improvements against a measurable objective.

What Remains

While the resource score has plummeted, the physical hardware required to execute these circuits—fault-tolerant logical qubits—does not yet exist at this scale. The focus now shifts to whether these algorithmic optimizations can be further compressed and how quickly networks can implement the NIST-standardized replacements. The arXiv paper associated with the challenge noted further reductions in the resource score to 1.259 billion, with some low-width circuits reaching as few as 813 qubits.