Ethereum zkAPI Goes Live: Pay for AI Without Revealing Your Identity Using Zero-Knowledge Proofs

Ethereum zkAPI Brings Private AI Payments to Mainnet

The Ethereum Foundation has launched Ethereum zkAPI on mainnet, introducing a privacy-preserving protocol that enables users to pay for AI without revealing identity1

1

. Built in collaboration with the Open Anonymity Project, the system uses zero-knowledge proofs to separate payment credentials from user identities when accessing metered API access2

2

. The launch transforms a February 2025 proposal by Ethereum co-founder Vitalik Buterin and Ethereum Foundation researcher Davide Crapis into a working implementation that addresses mounting privacy concerns around AI usage.

How the Private Payment System for AI Models Works

Users deposit funds like ETH or USDC into an Ethereum vault contract, creating a private note that functions as digital cash only they can spend1

1

. When requesting AI services, local software generates a zero-knowledge proof demonstrating sufficient funds exist without revealing which deposit belongs to them. The system then issues short-lived API keys with predefined spending limits, typically capped at a dollar amount and stored only in device memory1

1

. Requests go directly to the AI provider with just the temporary key—no name, no payment details. When keys expire, providers issue signed usage receipts and users are charged for actual consumption rather than the cap. The payment layer sees spending amounts while the provider sees prompts, but neither can link the two together1

1

.

Privacy-Preserving API Payments Address Real Concerns

The Ethereum Foundation emphasized that prompts reveal deeply personal information about health, finances, and doubts1

1

. Privacy concerns in AI aren't theoretical—in May 2025, a federal court ordered OpenAI to preserve output logs, including deleted chats, in a copyright lawsuit brought by The New York Times and other publishers1

1

. This legal precedent underscores why separating payment from identity matters for users who want to maintain privacy while accessing AI services. The ZK API Usage Credits design provides an alternative to handing a running transcript of your thinking to whoever controls the billing relationship1

1

.

Beyond AI: Broader Applications for zkAPI

While AI represents the immediate use case, Ethereum Foundation documentation identifies multiple applications for this privacy-preserving protocol3

3

. The system can handle blockchain data queries, image and video generation, VPN bandwidth, and machine-to-machine payments. This architecture creates a payment route for AI agents that need to purchase services automatically without opening conventional accounts with every provider3

3

. Davide Crapis, who leads the Ethereum Foundation's dAI team, previously proposed the ERC-8004 standard in August 2025—a shared rulebook letting autonomous AI agents find each other, verify identity, and transact1

1

. Crapis predicted most Ethereum traffic would come from machines within three to five years, positioning zkAPI as critical infrastructure for that future.

Integration and Technical Implementation

The launch includes OA Chat, a browser-based private chatbot requiring no installation, with open-source code available on GitHub1

1

. The project's documentation points to OpenRouter as the AI provider behind the keys—a service offering access to hundreds of AI models through one connection1

1

. Any application supporting the OpenAI format can integrate by pointing at a local address. The local client exposes standard OpenAI and Ollama interfaces through localhost, allowing existing applications to add the privacy layer without rebuilding surrounding AI software3

3

. Ethereum handles the settlement layer through deposits, withdrawals, and balance closures on-chain, while routine spending proofs are verified off-chain to avoid transaction overhead for every API request3

3

.

Privacy Limitations Users Should Understand

Ethough zkAPI separates payment from identity, complete anonymity remains elusive. The system does not hide prompt contents or network metadata from providers, meaning users can potentially be linked across sessions through IP addresses, timing, or information contained in requests2

2

. Personal details, writing style, reused conversation history, and project documents in prompts may allow inference providers to associate separate sessions3

3

. The Ethereum Foundation acknowledges that stable network identifiers may enable gateways to correlate activity, separating payment privacy from network anonymity and content privacy3

3

. Users seeking complete anonymity would need to combine zkAPI with additional privacy tools.

Experimental Status and Future Integration

The project repository labels the protocol experimental and does not list a formal audit1

1

. Wider adoption depends on integration by developers, support from API providers, user demand, and system performance as usage scales3

3

. Providers would need to accept zero-knowledge proofs instead of standard API keys and settle signed usage receipts to integrate directly1

1

. The mainnet launch arrives as the Ethereum Foundation continues its broader 2026 engineering program focused on scaling, user experience, and Layer-1 improvements, including plans to move the gas limit toward and beyond 100 million3

3

. Watch whether major AI providers adopt the standard and whether the experimental protocol withstands real-world usage demands.