Ethereum Foundation Launches Privacy Tool That Cuts the Link Between AI Prompts and Payers
By Unchained
AI summary of the source article
The Ethereum Foundation's dAI team, alongside the Open Anonymity Project, has launched an experimental protocol called zkAPI on Ethereum mainnet. The system disconnects user payment identities from AI prompts by having users deposit crypto, such as ETH or USDC, into an onchain vault. Device software creates zero-knowledge proofs of funds to secure temporary, capped API keys from a zkAPI server, ensuring the billing server never sees prompt content and the AI provider never receives billing details. However, zkAPI does not provide network anonymity, meaning providers can still view IP addresses, request content, and potentially re-link sessions through repeated prompt details.
Why it matters
Standard AI API access ties user prompts directly to accounts and payment methods, exposing sensitive personal data like financial and health queries. zkAPI decouples the payment layer from inference requests so AI providers cannot trace prompts to a billing identity.
Key facts
- The Ethereum Foundation and Open Anonymity Project launched zkAPI as an experimental protocol running on Ethereum mainnet.
- Users deposit ETH or USDC into a vault contract and generate zero-knowledge proofs to obtain temporary, dollar-capped API keys.
- The tool does not conceal network metadata, meaning AI providers can still see request contents and user IP addresses.