CandyPulse

AI Knows Your Prompts. Does It Need Your Identity? Ethereum's zkAPI Explained

By · 2026-10-02 · Blog

Think about the last few things you asked an AI assistant.

Maybe it was a question about a health symptom. A draft of a difficult email. Advice on money, a relationship or a job you're thinking of leaving. Now imagine every one of those questions neatly filed under your name, email address and card number.

For most AI services today, that's roughly how it works. The service knows what you asked, and it knows exactly who you are.

The Ethereum Foundation thinks there's a better way. Its new system, zkAPI, explores a simple idea: an AI service needs your payment, not your identity.

Key takeaways

The privacy problem with AI

AI services usually need three things from you: an account, a payment method and your prompts. Put those together and you get a detailed, personal profile.

That data can be valuable to the provider, attractive to hackers and sensitive if it ever leaks. Even with good intentions, a company holding millions of identified conversations is a big privacy risk.

The key insight behind zkAPI is that the AI provider doesn't actually need to know who you are. It needs to know you've paid.

How zkAPI works, in plain English

zkAPI was launched by the Ethereum Foundation with the Open Anonymity Project, based on a design by Vitalik Buterin and the Foundation's Davide Crapis.

Think of it like a prepaid card bought with cash:

  1. Top up. You deposit ETH or USDC into a vault on Ethereum.
  2. Prove, don't reveal. When you use an AI service, your device creates a zero-knowledge proof: a mathematical receipt showing you have enough credit and haven't spent it already, without saying who you are.
  3. Get a key. The service checks the proof and gives you a temporary access key with a spending limit and an expiry time.

The result is a neat separation:

WhoWhat they learn
The AI providerYour request, but not your identity
The payment processorOnly the payment amount
YouFull access, without handing over your name or card

What zero-knowledge proofs make possible

A zero-knowledge proof lets you prove something is true without revealing the details. It's like proving you're over 18 without showing your name, address or exact birthday, just a "yes" that can be checked.

That's what makes zkAPI possible: proving "I've paid" without revealing "I am this person".

Payment privacy, not total anonymity

Here's the honest part, and the team is upfront about it. zkAPI protects who paid, but it can't hide everything:

So zkAPI isn't an invisibility cloak. It removes one big link between you and your AI use, the payment trail, but you still need to think about what you share. The project is also labelled experimental, so treat it as a promising early tool rather than a finished product.

Why this matters beyond AI chat

The biggest impact may come from AI agents: software that acts on your behalf, buying data, calling services and paying other agents. As agents become more common, they'll make huge numbers of small payments.

Those payments need to be fast, automatic and private. Nobody wants their AI assistant's every purchase broadcast publicly or tied to their identity.

That's why projects across crypto are building payment rails for AI. On CandyChain, for example, the upcoming AI Agent Marketplace is designed to let agents be hired and paid in CANDY, and CandyX402 is built for machine-to-machine payments. zkAPI adds a privacy layer to the same big idea: an economy where software pays software.

How to protect your AI privacy today

You don't have to wait for new technology to be more careful:

  1. Avoid sharing identifying details in prompts unless you need to.
  2. Check privacy settings, including whether your conversations are used for training.
  3. Use separate accounts for sensitive topics where possible.
  4. Assume prompts may be stored, and write accordingly.
  5. Watch for privacy-first tools like zkAPI as they mature.

The bottom line

AI already knows a lot about what you're thinking. The question zkAPI raises is whether it also needs to know who you are. By letting people pay without revealing their identity, Ethereum is testing a future where using AI doesn't mean handing over your personal life.

It's early, it's experimental, and it's not total anonymity. But it's a meaningful step towards a world where your questions can stay yours.

This article is for information only and is not financial advice.

More on CandyPulse

A Dollar Built on Kaspa? The KSD Whitepaper Outlines a New Path for Native DeFi

When AI Starts Paying: The Economy CandyChain Is Building Toward

The Future Doesn't Build Itself: How CANDY Builds, Releases, Improves and Repeats

You Set the Goal, Agents Do the Work: Why AI Agent Payments Will Shape the Next Internet, and How CANDY Is Building x402 On Chain

Trending now

One CANDY, More Ways to Use Crypto: Pay, Play and Create Across the CANDY Ecosystem

Create Tokens Without Code: What CandyChain Studio Could Mean for Creators, Games and Communities

From Playing Games to Everyday Spending: How CandyRush Rewards Connect to Cardaxo

Swipe. Earn CANDY: How Cardaxo Turns Everyday Spending Into Crypto Rewards

Nvidia Falls More Than 4% as Board Investigation Into CEO Allegations Weighs on Shares

Wall Street Roars Back: Dow Jumps 1,104 Points as Oil Slides and Chip Stocks Rebound