Ethereum Foundation puts anonymous pay-per-call AI billing on mainnet with zkAPI
Ethereum Foundation launched zkAPI on mainnet, letting users pay for AI model calls from a private vault without revealing who is paying. The project's own repository still labels the protocol experimental, so early use will come from privacy-minded users and AI agents willing to test it.
The Investor · Invest desk

What happened
- Users deposit ETH or USDC into a vault contract on Ethereum that records each balance as a private note, then prove a request is funded without showing which note pays for it.
- A zkAPI server checks that proof, issues a temporary API key with a spending cap, and deducts the usage from the user's private balance once the key expires.
- The system implements ZK API Usage Credits, a design Vitalik Buterin and Davide Crapis posted to the Ethereum Research forum on February 11, and was built with the Open Anonymity Project.
- The post lists AI chat and agents as the first use case, followed by blockchain RPC queries, image and video generation, VPN bandwidth and machine-to-machine payments between agents.
Why it matters
- What it limits constraint Payment privacy only holds for users who also route traffic through Tor and strip identifying detail from prompts. That setup cost limits the early user base to people who already care about anonymity.
- What's now possible capability Developers can try it without rewriting their tools, because the local client mimics the standard OpenAI and Ollama APIs and existing apps only need to point at localhost.
- What it sets up precedent The dAI team, formed in September 2025, has now shipped two mainnet products since January (ERC-8004 and zkAPI), so designs its leads post to the research forum can reach mainnet within months.
Next story · Invest · 3h ago
a16z's Aerie challenges Ethereum's hash-based quantum plan with lattice signaturesConfirmed2 publishers
In credit terms, the zkAPI server lends. According to the Foundation's post, it authorises spending up to a cap before it settles, and it settles against a balance it has been shown is funded but cannot tie to a payer [2][5]. The payment layer never handles the work itself, because prompts go directly to the model provider [2]. A published serial number stops double-spending. "A user who tries to spend the same balance twice produces a duplicate nullifier, which exposes the attempt and nothing else," said the post, written by Vittorio Rivabella of the Foundation's dAI team [8][15].
The launch materials as reported do not include a fee, the rule linking a key's cap to the note behind it, or how the model provider receiving those prompts gets paid.
"Every AI API call today carries an identity," the post said. "Your API key points to an account, the account to a payment method, and every prompt you send joins the record attached to both." [9] zkAPI takes the account and the payment method out of that chain [5]. The prompt stays in it. The post concedes that sessions can still be re-linked through personal details, writing style or conversation history [7]. It also concedes that a gateway can correlate requests from a stable IP address, and it leaves that gap to Tor [6].
I see three ways this develops. In the first, it stays a tool for people who already route through Tor and keep small balances. In the second, a commercial model provider accepts the keys as a billing channel and brings ordinary chat users with it. In the third, the volume comes from software paying software, meaning the agent-to-agent and RPC uses on the post's list [11]. "It's like a generalization of OA unlinkable inference that also abstracts payments away," Ken Liu, a Stanford computer science PhD candidate who works on the Open Anonymity Project, said in an X post [16].
I'd expect the third path to produce volume first. An RPC query or a payment between two agents carries little of the personal detail, writing style or conversation history that the post says can re-link sessions [7]. That avoids one of the two limits the post names. The other limit is the counter-case: an agent running from a fixed server address is exactly the stable-IP correlation the post describes [6]. If the vault holds little ETH or USDC six months from now [1], none of the three paths has started and the view is wrong.
What to watch
- A commercial AI model provider saying it accepts zkAPI keys, which would test whether ordinary chat users can be reached.
- The GitHub repository dropping its experimental label for the protocol.
- Any change to the protocol that adds network anonymity, so it no longer depends on users routing through Tor.
Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence68
- Adoption10
- Hype gap+15
- Incentives60
- Confidence72
Perspective Coverage
11 publishers- Builder
- Builder 41%
- Operator
- Operator 36%
- Investor
- Investor 23%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Users first deposit tokens such as ETH or USDC into a vault contract on Ethereum, which records the balance as a private note; software on the user's device generates a zero-knowledge proof showing the request is backed by a funded note without revealing which note or deposit funds it.
- [2]
A zkAPI server checks the zero-knowledge proof and issues a temporary API key with a spending cap; the user sends prompts directly to the AI model provider with that key, and when it expires the server deducts the usage from the user's private balance.
- [3]
The Ethereum Foundation on Thursday launched zkAPI, a system that lets users pay for AI models and other metered APIs without revealing who they are, according to a Foundation blog post.
- [4]
The Foundation built zkAPI with the Open Anonymity Project, an open-source AI privacy effort, and it is live on Ethereum mainnet.
- [5]
Users authorize spending with zero-knowledge proofs, so the API provider and the payment server cannot link requests to a payer.
- [6]
zkAPI does not provide network anonymity; the gateway can potentially correlate requests sent from a stable IP address, and users seeking stronger privacy may route traffic through Tor.
- [7]
Sessions can also be re-linked through the content of prompts, such as personal details, writing style or conversation history.
- [8]
"A user who tries to spend the same balance twice produces a duplicate nullifier, which exposes the attempt and nothing else."
- [9]
"Every AI API call today carries an identity. Your API key points to an account, the account to a payment method, and every prompt you send joins the record attached to both. The provider can connect years of your usage into a single profile."
- [10]
"The client exposes the standard OpenAI and Ollama APIs on your own machine. Existing apps, editors, and chat clients work by pointing them at localhost."
- [11]
The post lists AI chat and agents as the first use case; other listed applications include blockchain RPC queries, image and video generation, VPN bandwidth and machine-to-machine payments between AI agents.
- [12]
The zkAPI project's GitHub repository describes the protocol as experimental.
- [13]
zkAPI implements 'ZK API Usage Credits,' a design Davide Crapis and Ethereum co-founder Vitalik Buterin published on the Ethereum Research forum on Feb. 11.
- [14]
Crapis leads the Ethereum Foundation's dAI team, formed in September 2025 to improve Ethereum as a settlement and coordination layer for AI; the team also developed ERC-8004, an AI agent identity standard that launched on mainnet in January.
- [15]
The blog post announcing zkAPI was written by Vittorio Rivabella of the dAI team.
- [16]
"It's like a generalization of OA unlinkable inference that also abstracts payments away."
Sources
11 independent publishers whose own reporting we read for this story.
- bankless.comEthereum Foundation Launches zkAPI for Private AI Payments
1 article · October 2, 2026
- cointelegraph.comEthereum’s zkAPI brings privacy-preserving API payments to mainnet
1 article · October 1, 2026
- crowdfundinsider.comEthereum Foundation and Open Anonymity Launch zkAPI for Private Metered API Payments on Mainnet
1 article · October 2, 2026
- crypto.newsEthereum’s zkAPI hides who pays for AI. The model still sees the prompt
1 article · October 2, 2026
- cryptobriefing.comEthereum Foundation and Open Anonymity launch zkAPI for private AI access
1 article · October 1, 2026
- cryptopolitan.comEthereum targets AI privacy with zkAPI for anonymous API payments
1 article · October 1, 2026
- cryptoslate.comVitalik-inspired AI payment system can send expired deposits to its treasury
1 article · October 4, 2026
- decrypt.coEthereum Now Lets You Pay for AI Without Revealing Who You Are
1 article · October 2, 2026
- dev.toWhat is Ethereum zkAPI? A zero-knowledge proof is not a payment authorization
1 article · October 2, 2026
- news.bitcoin.comEthereum’s zkAPI Gives AI Payments a Privacy Cloak With a Catch
1 article · October 4, 2026
- theblock.coEthereum Foundation launches zkAPI to let users pay for AI models without revealing identity
1 article · October 1, 2026
Topics and entities
Entities
- Vitalik Buterin
- Vittorio Rivabella
- Open Anonymity Project
- OpenRouter
- Ethereum
- Ollama
- Human Rights Foundation
- Ethereum Foundation
- Tor