AI Agents Just Slashed the Cost of a Quantum Attack on Bitcoin
Researchers using AI coding agents cut a key quantum-attack resource benchmark for cracking secp256k1, the elliptic-curve cryptography behind Bitcoin and Ethereum signatures, by 86%. The ECDSA.Fail competition, launched by Eigen Labs, drew 100+ participants and produced a circuit that verified the needed calculations but did not recover any private key. The benchmark measures logical qubits multiplied by Toffoli gates; the best design reduced the score from 10.75 billion to 1.496 billion, using 1,151 logical qubits and about 1.3 million Toffoli gates. The work suggests one step in a future “Q-Day” attack may be easier than previously estimated, even though the timeline for capable quantum computers remains uncertain. The researchers also framed the project as a model of human-AI “open autoresearch,” while crypto firms are increasing funding for post-quantum and Bitcoin security research.
