Eigen Labs and Shielded Labs launch HashSmash, a competitive program testing AI-based attacks on cryptographic hash functions.
AI & Agents ·
Eigen Labs and Shielded Labs have launched HashSmash, an open competition to test AI-based attacks on cryptographic hash functions. The project, announced in early October 2026, focuses on evaluating the security of widely-used algorithms including SHA-256, SHA-3, and BLAKE3, as well as post-quantum variants. Because the cryptographic security of these functions remains unproven, the initiative invites community participation to discover potential vulnerabilities through competitive submissions tracked on public leaderboards.
The platform currently hosts multiple hash challenges with varying submission volumes—SHA-256 has received 6 submissions, SHA-3 256 has 7, and BLAKE3 has 9, with additional challenges planned. Submissions are reviewed by a panel of experts including JP Aumasson and Dmitry Khovratovich, who co-designed some of the tested algorithms. Results published on the leaderboard display attack parameters such as round counts and computational complexity metrics, though entries remain subject to manual review before formal acceptance.
The extent to which AI-assisted methods will advance attack discovery beyond conventional cryptanalysis, and whether any published results will lead to practical changes in hash function recommendations, remains to be determined as submissions continue to accumulate.