Overview
- A collaboration involving over 100 participants utilized AI agents to decrease the resource score for a quantum attack from 10.75 billion to 1.496 billion.
- The top-performing circuit employed 1,151 logical qubits and approximately 1.3 million Toffoli gates, although comparisons with Google’s findings should be approached with caution.
- A subsequent design successfully lowered the gate count to under 1 million; however, this does not factor in the hardware expenses associated with a complete attack.
A recent study revealed that researchers using AI coding agents have managed to cut the resource benchmark required for a critical phase in a potential quantum attack on Bitcoin and Ethereum's cryptographic systems by 86%. This research was detailed in a paper released on Wednesday.
This achievement means that the computing resources necessary to expose a wallet's private key—potentially allowing theft of its assets—have been significantly reduced. The timeline for the advent of sufficiently powerful quantum computers capable of executing such an attack, a moment referred to as “Q-Day”, remains unknown.
Myriad: What will Bitcoin's next move be? Make your prediction here.The researchers noted, “While the timing is still uncertain, the shift away from vulnerable cryptography has already begun.” They highlighted that NIST has standardized alternatives to current cryptographic methods, with an initial public draft of NIST IR 8547 suggesting that classical public-key algorithms at the 112-bit security level could be phased out after 2030 and banned entirely by 2035.
The insights were derived from the ECDSA.Fail competition, initiated by Eigen Labs in late May, where more than 100 participants developed circuits for secp256k1, the elliptic curve responsible for securing transaction signatures on Bitcoin and Ethereum. Notable contributors to the paper include researchers from Theta Labs, MultiVM Labs, Eigen Labs, Trail of Bits, StarkWare, and the Ethereum Foundation.
The researchers successfully designed and evaluated a quantum circuit capable of executing calculations necessary to breach Bitcoin’s cryptography, achieving an impressive 86% reduction in resource requirements. However, these tests confirmed the circuit's calculations without actually cracking a Bitcoin private key.
The challenge evaluated the quantum memory and computational effort required by each circuit, calculating the resource score by multiplying the number of logical qubits by the Toffoli gates used, which are a complex type of quantum operation developed by Tommaso Toffoli in 1980. A lower score indicates that fewer resources are needed for the calculation, potentially facilitating a future attack.
By July 26, participants had managed to slash the resource score from 10.75 billion to 1.496 billion. The leading design utilized 1,151 logical qubits and around 1.3 million Toffoli gates. As per the study, this score was roughly half of what Google Quantum AI achieved in March, though direct comparisons are complicated by differences in methodologies.
The researchers concluded, “In addition to the resulting circuits, ECDSA.Fail exemplifies a case study of Open Autoresearch: a verifier-gated research process where human participants and AI agents collaboratively generate, implement, test, and share potential enhancements against a common measurable goal.”
This report is part of an increasing trend among cryptocurrency firms to invest in research aimed at countering quantum threats.
In July, Galaxy Digital pledged up to $5 million to advance this research, while nine companies, including BlackRock, Coinbase, and Strategy, committed a total of $15 million over three years for broader Bitcoin security research, which includes defenses against quantum attacks.