HashSmash Competition Tests AI Attacks on Core Cryptographic Hash Functions

Two prominent organizations in the blockchain and cryptography space have joined forces to probe the resilience of the hash functions that secure much of the modern internet. Eigen Labs and Shielded Labs, the organization behind the Zcash privacy protocol, have launched HashSmash, an open multiplayer competition hosted on the Yukon platform that invites participants to deploy artificial intelligence agents against widely used cryptographic primitives.

The competition targets four hash functions — SHA-256, SHA-3, BLAKE3, and Poseidon — that collectively underpin encryption, digital signatures, and data integrity checks across countless applications, from web browsing to blockchain networks. By opening the challenge to anyone, the organizers aim to crowd-source adversarial pressure on these algorithms at a scale that conventional academic research rarely achieves.

Read more FDA Clears Second Alvotech Plant for Humira Biosimilar Production →

A key feature of the initiative is its transparency. Every result produced by competing AI agents will be published and reviewed by professional cryptographers, creating a continuously updated, publicly auditable record of how close machine-learning techniques come to weakening established hash constructions. That open-review model is intended to surface vulnerabilities before bad actors can exploit them privately.

The initiative reflects a broader anxiety in the security community about whether AI-accelerated computational power could eventually erode confidence in cryptographic standards that were designed well before modern large-scale machine learning existed. By stress-testing these functions now, organizers say the goal is to give the industry time to adapt if meaningful weaknesses are discovered.

Continue reading at Cryptocurrency