Ethereum Foundation Launches zkAPI: Pay for AI Models Using Zero-Knowledge Proofs Without Exposing Identity

Compiled by TechFlow

TechFlow Summary: The Ethereum Foundation and the Open Anonymity Project have turned Vitalik et al.'s February proposal for "ZK API Usage Credits" into a mainnet-ready zkAPI: users deposit ETH/USDC into a vault first, then use zero-knowledge proofs to exchange for quota-bearing temporary keys to call models, separating payment from identity within the API chain. This adds to narratives around the ETH settlement layer and AI agent machine-to-machine payments; however, gateways may still correlate requests via IP, and the protocol itself is still marked as experimental.

The system was built with the participation of the Open Anonymity Project, implementing a design publicly released in February by Ethereum co-founder Vitalik Buterin and Davide Crapis, head of the Ethereum Foundation's dAI team.

In a blog post published Thursday, the Ethereum Foundation announced the launch of zkAPI—a system that allows users to pay for AI models and other usage-based APIs without exposing their identity.

The Foundation stated that zkAPI was developed jointly with the open-source AI privacy project Open Anonymity (OA) Project and is now live on the Ethereum mainnet.

"Every AI API call today carries an identity," the blog wrote. "Your API key points to an account, which in turn points to a payment method. Every prompt you send gets appended to logs attached to both. Service providers can stitch together your years of usage history into a single profile."

"It feels like a generalization of OA's unlinkable inference, while also abstracting away payment," Stanford computer science PhD candidate Ken Liu, who participates in the Open Anonymity Project, posted on X Thursday afternoon. He added, "There should be more infrastructure that takes privacy and sovereignty as its foundation, allowing people to truly control themselves."

How zkAPI Works

According to the blog, when using zkAPI, users must first deposit tokens such as ETH or USDC into a vault contract on Ethereum, which records user balances as private vouchers. To spend these tokens, software on the user's device generates a zero-knowledge proof demonstrating that their request is backed by a voucher with sufficient funds, without revealing which specific voucher it is or exposing the corresponding deposit.

After verifying the zero-knowledge proof, the zkAPI server issues a temporary API key with a spending cap. Users then send prompts directly to the AI model provider using this key; when the key expires, the server deducts the corresponding usage from the user's private balance.

Each time a user pays for AI usage, they also publish a serial number called a "nullifier." "If someone attempts to spend the same balance twice, a duplicate nullifier will be generated, exposing the attempt without leaking any other information," the blog noted.

According to the blog, the client can work alongside existing AI tools. "The client exposes standard OpenAI and Ollama APIs locally on your machine," the blog stated. "Existing apps, editors, and chat clients only need to point to localhost to function."

Use Cases and Limitations

The blog lists AI chat and agents as primary use cases. Other highlighted applications include blockchain RPC queries, image and video generation, VPN bandwidth, and machine-to-machine payments between AI agents.

However, according to the blog, the zkAPI system does not provide network-layer anonymity. Gateways may correlate requests originating from fixed IP addresses; for stronger privacy, the blog suggests routing traffic through Tor—a decentralized anonymous network that relays traffic through multiple servers and is also commonly used by so-called "dark web" sites.

Sessions could also be re-correlated through prompt content, such as personal details, writing style, or conversation history. The zkAPI project's GitHub repository describes the protocol as experimental.

Design Co-authored by Vitalik Buterin

zkAPI implements the "ZK API Usage Credits" design, which was published by Davide Crapis and Ethereum co-founder Vitalik Buterin on the Ethereum Research forum on February 11.

Crapis leads the Ethereum Foundation's dAI team, which was established in September 2025 to enhance Ethereum's capabilities as an AI settlement and coordination layer. The team also developed the AI agent identity standard ERC-8004, which went live on the mainnet in January.

The blog announcing zkAPI was written by Vittorio Rivabella of the dAI team.

In a February X post, Buterin listed cryptographic payment mechanisms for AI services as one of Ethereum's near-to-medium-term visions regarding its role in AI.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps