Ethereum Foundation Launches zkAPI for Private AI Payments

The Ethereum Foundation and the Open Anonymity Project launched zkAPI on Ethereum mainnet on October 1, 2026, a system that lets users pay for metered APIs without revealing who is paying, according to a post on the Ethereum Foundation blog. The first target is AI inference, where prompts are sensitive and the billing trail currently ties every request to an identity.

Separating Payment From Identity

Users deposit credits such as ETH or USDC into a vault contract on Ethereum in a single transaction. From that point the balance exists as a private note that only the holder can spend. Software on the user’s machine produces a zero-knowledge proof showing that a funded note covers the spend and has not been spent before. Every spend publishes a nullifier, a one-way serial number that exposes double-spending attempts without revealing which note backed the payment.

AI Inference Is the First Use Case

zkAPI ships a local client that exposes the standard OpenAI and Ollama APIs, so existing apps and chat clients keep working by pointing at localhost. A server checks a payment proof, mints a short-lived API key held only in the device’s memory, and charges the private balance when the key expires. The foundation describes a split where the payment server sees that a valid payment exists but not who made it, while the AI provider sees prompts but not the payer. The same client can front RPC queries, image generation, VPNs, and machine-to-machine services, making it relevant beyond AI agents in crypto.

A Design Co-Authored by Vitalik Buterin

The launch is the working implementation of ZK API usage credits, a design published on Ethereum Research by Davide Crapis and Vitalik Buterin. It proves spends with Groth16 proofs on the BN254 curve, hashes commitments and nullifiers with Poseidon, and keeps notes in a 32-level Merkle tree. The live vault contract on mainnet holds USDC credits, and withdrawals stay available onchain even if the zkAPI server disappears.

Limits and Next Steps

The foundation notes two limits: zkAPI does not provide network anonymity, so users seeking stronger privacy should route through Tor, and shared prompt contents can still act as fingerprints. The project is under active development, with confidential GPU computing still emerging for content privacy. The launch extends Ethereum into private AI infrastructure, where billing records currently link every query to a payer.