Ethereum plans to move neural network computations off-chain, creating a trusted environment for autonomous programs. Davide Crapis, head of AI at the Ethereum Foundation, outlined this approach at NEARCON 2026. The goal is to protect core crypto values—user sovereignty, privacy, and censorship resistance—by avoiding dependence on centralized AI. Ethereum does not intend to compete with OpenAI or Google in language models, focusing instead on preventing re-centralization of internet infrastructure.
The strategy unfolds in two directions. First, DECENTRALISED AI COORDINATION envisions an on-chain registry for identifying agents, enabling payments and reputation building while keeping heavy computations off-chain on traditional servers. The blockchain acts as a public directory to help agents find each other and transfer funds, likened to a decentralised version of Google reviews with built-in payments. The second direction, Props AI, explores processing data locally on users’ devices to avoid surveillance, while a standardised AI agent identification protocol—ERC-8004—gains traction beyond Ethereum.
Crapis warned of rising cyber threats as neural networks may learn to impersonate humans, potentially undermining traditional passwords. He suggested the security model should rely on cryptographic keys with mathematical verification capabilities. Integrating AI has become a priority for the Ethereum Foundation, with developers expecting algorithms to shape a new network era and make the blockchain a fundamental layer for smart agents. Back in February, project co-founder Vitalik Buterin extensively analysed potential scenarios for synergy between Ethereum and artificial intelligence.
Leave a Reply