# Ethereum Foundation's zkAPI Separates AI Payments From User Identity

> On October 1, 2026, the Ethereum Foundation introduced zkAPI, an open-source system built with the Open Anonymity Project that lets users pay for AI services from an Ethereum vault using zero-knowledge proofs. The AI provider sees prompts but not who paid, and the payment server sees spending but not the prompts. The foundation was clear that it doesn't hide network traffic or what users reveal in their own messages.

Source: https://cryptomorrow.com/2026/10/01/ethereum-foundations-zkapi-separates-ai-payments-from-user-identity/
Last checked: 2026-11-01

Paying for an AI service normally leaves a durable trail linking an account, a credit card or wallet, and a history of prompts. On October 1, 2026, the [Ethereum Foundation](https://ethereum.foundation) introduced zkAPI, a working open-source system designed to break the link between what an API customer pays and who they are. It was built by the [Open Anonymity Project](https://openanonymity.ai) with the foundation, and the [announcement](https://blog.ethereum.org/en/2026/10/01/introducing-zkapi) said it already runs on Ethereum mainnet.

*Historical news event: 2026-10-01. Retrospective report prepared November 2026.*

## The problem zkAPI is trying to solve

"Every AI API call today carries an identity," Vittorio Rivabella of the foundation's dAI Team wrote in the October 1 post. An API key points to an account, the account points to a payment method, and every prompt joins the record attached to both. Over time, a provider can build a single profile of years of usage, and prompts are often personal: people ask chatbots about their health, money and private worries.

The design wasn't new on launch day. In February 2026, Davide Crapis, the Ethereum Foundation research scientist who leads its dAI Team, and Ethereum co-founder [Vitalik Buterin](/2017/11/11/vitalik-buterin/) published a proposal for anonymous API usage credits on the Ethereum Research forum. "We need a system where a user can deposit funds once and make thousands of API calls anonymously, securely, and efficiently," they [wrote](https://ethresear.ch/t/zk-api-usage-credits-llms-and-beyond/24104) on February 11. zkAPI is the working version of that idea, with the Open Anonymity team building the client, server and smart contracts.

![Portrait of Ethereum co-founder Vitalik Buterin in a white T-shirt in front of a patterned wall](/static/up/2026/10/cm-003-vitalik-buterin-portrait.webp)

*Photo: Romanpoet / Wikimedia Commons, CC BY-SA 4.0*

## How the privacy model works

The system relies on [zero-knowledge proofs](/2018/10/31/zkp-technology-to-let-enterprises-do-private-transactions-on-public-blockchains/), a cryptographic technique that proves a fact without revealing the secret behind it. Here, a customer proves they have enough prepaid credit for a request without revealing which deposit funded it.

A user first deposits funds into a vault contract on [Ethereum](/2017/08/21/ethereum-eth/) in a single transaction; the post says the vault holds USDC credits. That balance becomes a private note that only the depositor can spend. Software on the user's device then creates a proof that a funded note covers the spend and hasn't been used before. The zkAPI server checks the proof and issues a short-lived API key with a dollar cap, held only in the device's memory.

The user's prompts then go straight from the device to the AI provider, not through the payment service. When the key expires, the provider signs a usage receipt and the server deducts the actual amount from the private note. To stop the same balance being spent twice, each spend publishes a one-way serial number called a nullifier; reusing a note produces a duplicate and exposes the attempt.

"The provider sees the requests, and the payment layer sees the spend. Neither learns the link between them," the post said. According to the foundation, the zkAPI server knows a valid payment exists and how much was spent per session, but not the user's identity, the requests or which deposit paid. The AI provider sees prompts and responses but not who's paying. The public blockchain shows deposits and withdrawals, but not what any balance paid for.

Funds also aren't held by a company. "The vault is a contract on Ethereum rather than a company account," the post said, adding that users can close their balance and withdraw onchain even if every zkAPI server disappears.

## Built to fit existing tools

The local client exposes the standard OpenAI and Ollama programming interfaces on the user's own machine, so existing apps, editors and chat clients can connect by pointing to a local address instead of adopting a new workflow. Providers accept proofs instead of API keys and settle signed receipts instead of managing customer accounts.

The code is public in the [zkAPI repository on GitHub](https://github.com/ethereum/zkapi), and the Open Anonymity Project runs a [browser chat demo](https://chat.openanonymity.ai) and [documentation](https://zkapi.openanonymity.ai). There's also a simpler "proxy mode," in which the zkAPI server relays requests itself. The foundation noted that in that mode the relay can see the traffic, which is why the direct key mode exists.

## Private billing isn't private thought

The foundation was careful about the limits. "zkAPI's protections have two limitations: the lack of network anonymity, and privacy leakage via prompt contents," the post said. A gateway can correlate request patterns from a stable IP address, so it suggests users who want stronger privacy route traffic through Tor with a fresh circuit for each session.

The bigger limit is the content itself. The AI model still sees the actual message and response, because it needs them to do its job. Repeated personal details, a recognizable writing style, reused chat history or shared documents can let a provider re-link sessions. The post called this a tradeoff between privacy and usefulness, and pointed to local models or models running in secure hardware enclaves as ways to reduce it. Paying privately, in other words, doesn't make every request anonymous.

## Uses beyond chatbots

The post listed AI chat and AI agents first, followed by blockchain data queries, image and video generation, VPNs and bandwidth, and machine-to-machine services where software agents pay per task without opening an account. That last case puts zkAPI alongside other 2026 efforts to let AI agents pay for services, such as [Solana and Google Cloud's Pay.sh](/2026/05/05/solana-and-google-cloud-introduce-pay-sh-for-ai-agents-to-pay-for-apis/), though zkAPI's focus is privacy rather than payment rails. It's also part of a wider moment for privacy tools in crypto that included [Grayscale's Zcash ETF](/2026/08/25/grayscales-zcash-etf-starts-trading-on-nyse-arca-bringing-privacy-coin-to-brokerages/) reaching brokerages.

The significance is as much about ideas as code. Crypto infrastructure has often made payments public by default. zkAPI shows an alternative: let people prove they've paid while revealing less about what they do. It comes with clear privacy benefits and clearly stated limits, and it's still early: a mainnet deployment on launch day isn't the same as a service proven at commercial scale.

## Frequently asked questions

### What is zkAPI?

zkAPI is an open-source system from the Open Anonymity Project and the Ethereum Foundation, announced on October 1, 2026. It lets users prepay for AI and other metered APIs from an Ethereum vault and spend those credits with zero-knowledge proofs, so usage isn't tied to their identity.

### Does zkAPI hide my prompts from the AI provider?

No. The AI provider still sees prompts and responses, because it needs them to run the model. zkAPI hides who paid for a request, not what the request says.

### Does zkAPI work with existing AI apps?

Yes, according to the foundation. The client runs on the user's machine and exposes standard OpenAI and Ollama interfaces, so existing apps can point to it locally.

### Can users get their money back if zkAPI servers shut down?

The foundation says yes. The vault is a smart contract on Ethereum, and users can close their balance and withdraw onchain even if every zkAPI server disappears.

### Who designed zkAPI?

The underlying design was published by Davide Crapis and Vitalik Buterin on the Ethereum Research forum in February 2026. The Open Anonymity Project built the working system with the Ethereum Foundation.

## Sources

- [Ethereum Foundation Blog: Introducing zkAPI, private usage credits for any API (October 1, 2026)](https://blog.ethereum.org/en/2026/10/01/introducing-zkapi)
- [Ethereum Research: ZK API usage credits, LLMs and beyond, by Davide Crapis and Vitalik Buterin (February 11, 2026)](https://ethresear.ch/t/zk-api-usage-credits-llms-and-beyond/24104)
- [GitHub: ethereum/zkapi](https://github.com/ethereum/zkapi)
- [Open Anonymity: zkAPI documentation](https://zkapi.openanonymity.ai)
- [Etherscan: ZkApiVault contract on Ethereum mainnet](https://etherscan.io/address/0x4386fdbda35d995beb3bf8625118ec5982ec81fe)

*Featured image: [Ethereum Foundation](https://blog.ethereum.org/en/2026/10/01/introducing-zkapi)*