More than 100 researchers working alongside AI coding agents significantly decreased a benchmark for a key step in a quantum attack on Bitcoin and Ethereum. The score dropped from 10.75 billion to 1.496 billion between late May and July 26, with the paper published this week. This development occurred within the ECDSA.Fail competition run by Eigen Labs, which scores circuit designs based on logical qubits and Toffoli gates.
The leading design used 1,151 logical qubits and roughly 1.3 million Toffoli gates, while a later submission pushed gates below a million. Authors include researchers from Eigen Labs, Trail of Bits, StarkWare, Theta Labs, MultiVM Labs, and the Ethereum Foundation. They utilized a method called Open Autoresearch, where humans and AI agents iterate against a shared measurable target. This work lands amid a broader industry scramble, including the Ethereum Foundation’s December 2029 deadline for quantum-resistant transactions and NIST’s draft proposing deprecation of classical public-key algorithms after 2030.
This development highlights a critical asymmetry in the race between cryptographic offense and defense. While institutional players have established multi-year roadmaps for upgrading infrastructure to quantum-resistant standards, the cost of attacking current elliptic curve cryptography has plummeted rapidly due to AI-assisted optimization. The reduction of the attack benchmark by 86% in just two months suggests that computational efficiency gains are accelerating beyond human-only research capabilities, potentially compressing the window available for secure migration.
Stakeholders must reassess risk models that assume attack research progresses at traditional speeds. The involvement of major security firms and foundations in publishing these findings indicates a proactive approach to transparency, yet it also underscores the urgency of the timeline. With NIST guidelines targeting the deprecation of classical algorithms by 2035 and specific deadlines like Ethereum’s 2029 mandate, the rapid advancement of AI-driven cryptanalysis serves as a stark reminder that defensive preparations may need to be accelerated to maintain network integrity.
