Ethereum Foundation launches zkAPI for anonymous AI payments

Oct 2, 2026

On October 1, 2026, the Ethereum Foundation, in collaboration with the Open Anonymity Project, launched zkAPI on the Ethereum mainnet. This experimental protocol allows users to pay for AI models and metered APIs anonymously using ETH or USDC. By utilizing locally generated zero-knowledge proofs, the system decouples payment credentials from user identities, preventing providers from linking prompts to specific payers. However, the protocol does not hide network-level metadata like IP addresses, and prompt content can still potentially reveal user identity.

Launch of zkAPI

▪The official GitHub repository for zkAPI labels the newly launched protocol as experimental.
▪The Ethereum Foundation launched zkAPI on the Ethereum mainnet on October 1, 2026, to enable anonymous payments for AI models and metered APIs.

Development and design origins

▪The Ethereum Foundation co-developed zkAPI with the Open Anonymity Project, an organization focused on open-source privacy solutions for artificial intelligence.
▪The zkAPI system implements a design co-authored by Ethereum co-founder Vitalik Buterin and Davide Crapis, director of the Ethereum Foundation's dAI initiative, published in February 2026.

Vault deposits and withdrawals

▪To use zkAPI, users must first deposit cryptocurrency, specifically ETH or USDC, into an Ethereum vault smart contract, which generates private notes on the user's local device.
▪Users can withdraw their leftover funds onchain from the vault contract even if every zkAPI server goes offline.

Cryptographic verification process

▪To prevent double-spending, zkAPI publishes a unique cryptographic identifier called a nullifier for each transaction, which exposes duplicate spending attempts while maintaining user anonymity.
▪The zkAPI protocol uses locally generated zero-knowledge proofs to verify that a user has sufficient funds in a private note without disclosing which specific deposit belongs to the requester.
▪A zkAPI gateway validates zero-knowledge proofs and issues temporary API keys with predetermined spending caps, allowing users to interact directly with AI providers.

Application integration

▪The zkAPI client software emulates standard OpenAI and Ollama API protocols locally, allowing existing applications to connect through localhost configurations without modification.
▪The Open Anonymity Project integrated zkAPI as a payment option within its OA chat application.

Anonymity limitations

▪The zkAPI protocol does not provide network-level anonymity, meaning gateway operators can still see IP addresses and metadata, and providers can re-link sessions through prompt content.
▪The Ethereum Foundation recommended that users seeking stronger privacy route their zkAPI traffic through the Tor anonymity network.

Privacy risks of traditional AI APIs

▪Traditional AI API interactions embed user identity into every request by linking API keys directly to user accounts and payment instruments, allowing providers to build comprehensive behavioral profiles.
▪AI interactions often contain highly sensitive personal details, business strategies, proprietary code, and financial information, making persistent account tracking a significant privacy risk.

Debatable claims

▪Ethereum is an appropriate settlement layer for the AI economy
▪zkAPI's privacy protections are too limited to secure sensitive AI interactions
▪Deploying experimental protocols like zkAPI directly to the mainnet is irresponsible
▪AI service providers should integrate anonymous payment protocols like zkAPI