The Ethereum Foundation has launched zkAPI, a new system designed to let users pay for AI models and other metered APIs without linking their usage to their payment identity.
Built with the Open Anonymity Project, zkAPI went live on Ethereum Mainnet on October 1. The system uses zero-knowledge proofs to separate payment information from API usage.
What is zkAPI?
Traditional AI APIs generally require users to connect their requests to an API key, account or payment method. This can allow providers to build a record connecting a user’s usage over time.
zkAPI takes a different approach.
Users deposit assets such as ETH or USDC into an Ethereum vault. The funds are represented as a private note, and users can later prove that they have funds available without revealing which deposit is being used.
The zkAPI server verifies the zero-knowledge proof and issues a temporary API key with a spending limit. The user’s prompts are then sent directly to the AI provider.
What does zkAPI hide?
The main goal is to separate who pays from what is being requested.
The payment server can verify that a valid payment exists, but does not receive the user’s prompts or the identity behind the payment.
Meanwhile, the AI provider receives the prompts needed to generate a response but is not given the billing identity associated with the payment.
This distinction is important: zkAPI does not make AI conversations completely anonymous.
The AI provider can still see the prompt, and network information such as an IP address can potentially be used to connect activity to a user. Repeated personal information or writing patterns in prompts can also make separate sessions identifiable.
Why Ethereum is involved
zkAPI uses Ethereum as the settlement layer for the payment system.
Users make a deposit into the vault, while zero-knowledge proofs allow subsequent spending to be authorized without revealing the specific deposit behind each payment.
The system also uses temporary API keys and spending limits, allowing users to interact with existing AI applications without changing the underlying model interface. The Ethereum Foundation says the client supports standard OpenAI and Ollama APIs.
Beyond AI
AI inference is the first major use case, but zkAPI is designed for other metered services as well.
Potential applications include blockchain RPC services, image and video generation, VPN bandwidth and payments between AI agents.
That could make the technology relevant as more software agents begin making automated payments for computing resources and APIs.
Built from earlier Ethereum research
The system is based on a design for zero-knowledge API usage credits published by Ethereum Foundation dAI lead Davide Crapis and Ethereum co-founder Vitalik Buterin earlier this year.
The launch moves that research from a design concept into a working system deployed on Ethereum Mainnet.
What comes next?
zkAPI is still an experimental system, and privacy at the payment layer does not automatically guarantee complete anonymity.
Its significance will likely depend on how developers integrate it, how the infrastructure performs under real-world use and whether users adopt privacy-preserving payments for AI and other API-based services.
For now, zkAPI represents an interesting convergence of Ethereum, zero-knowledge proofs, AI infrastructure and digital privacy.
