Ethereum Foundation, Open Anonymity bring zero-knowledge payments to API usage
AI & Agents ·
A new onchain vault system lets developers pay for metered API calls without exposing who paid or which funds were used.
The Ethereum Foundation and the Open Anonymity Project have rolled out zkAPI, a payment layer designed to let users settle metered API bills while keeping their identity separate from the transaction, according to wublockchain.xyz. The system is already running on Ethereum mainnet.
Under the model, users load ETH, USDC, or other supported credits into an onchain vault, then use zero-knowledge proofs to authorize spending. Providers can confirm that payment has been made without learning the identity of the payer or tracing which specific deposit funded a given request. The protocol can generate temporary API keys with built-in spending limits, with final settlement calculated against actual consumption rather than a flat fee.
The initial rollout targets AI inference services, where metered, pay-per-call pricing is common, but the underlying design is described as adaptable to other usage-based services beyond AI. That flexibility suggests the payment layer could eventually be plugged into a wider range of metered digital products, though no specific expansion plans beyond AI inference have been detailed.
The launch comes with an explicit limitation: zkAPI shields payment identity but does not mask network-level or content-level activity. IP addresses, request timing patterns, and any identifying details embedded in prompts themselves can still potentially be linked back to a user, meaning the privacy guarantee is confined to the financial transaction rather than the full interaction.
The cluster currently shows two distinct sources covering the same announcement, pointing to initial corroboration of the mainnet launch. What remains unclear is how zkAPI will be adopted by specific AI inference providers, what usage volume the vault system is handling at launch, and whether additional privacy layers addressing the network and content gaps are planned.