The Ethereum Foundation said zkAPI, a payment system for AI services developed with Open Anonymity Project, reached Ethereum mainnet on Oct. 1. It separates payment checks from model requests while leaving prompts visible to providers.
The foundation connected the project to concerns about personal data as AI use expands, including censorship and surveillance. The design addresses a billing model that can connect AI use with a customer account and its payment record. zkAPI is designed to reduce that connection by letting the payment process check charges without identifying the deposit backing them.
Users place funds in a vault and authorize spending with a zero-knowledge proof, a cryptographic check that can confirm payment authorization without disclosing which deposit supports it. The server issues a short-lived API key with a dollar cap that reserves spending for the session. Once the key expires, the server deducts the provider’s measured usage.
The privacy feature has limits. The payment process can validate charges without reading prompt text, but the AI provider receives the prompts. The provider can also see a user’s IP address. The foundation says prompt details and the timing of requests may allow separate sessions to be linked.
zkAPI therefore limits the information available during payment processing, while leaving the AI provider able to view the content and network details accompanying a request. The Ethereum Foundation announcement describes the system and its privacy limits.
