HashSmash Contest Pits AI Agents Against Core Crypto Hash Functions

Two prominent blockchain research organizations have joined forces to probe one of cryptography's most consequential questions: can artificial intelligence meaningfully advance attacks on the hash functions that underpin nearly all internet security. Eigen Labs and Shielded Labs, the development organization behind the Zcash privacy protocol, launched HashSmash, an open multiplayer competition hosted on the Yukon platform.

The contest invites any participant to deploy AI agents against four widely used hash functions — SHA-256, SHA-3, BLAKE3 and Poseidon. These algorithms form foundational layers of encryption across banking, communications, blockchain networks and government systems, making their resilience a matter of broad public concern beyond any single industry.

Read more Gate Bioscience Widens Lilly Deal to Target Hard-to-Drug Proteins →

Organizers say every result generated through the competition will be published openly and subjected to review by working cryptographers. The transparent structure is designed to surface genuine vulnerabilities rather than produce proprietary research that benefits only select institutions, a design choice that reflects growing calls for open science in AI safety research.

The timing is deliberate. As AI models grow more capable of pattern recognition and exhaustive search across large problem spaces, security researchers have warned that cryptographic standards considered safe for decades may warrant reassessment. HashSmash represents a structured, controlled environment to gather empirical data on that risk rather than relying on theoretical projections alone.

Whether the competition surfaces actionable weaknesses or confirms existing confidence in these hash functions, organizers argue the exercise advances the field by establishing a reproducible, community-driven benchmark for AI's cryptographic reach. Continue reading at Cryptocurrency.