AI News

Ethereum Foundation launches zkAPI for anonymous AI service payments

AI NewsCrypto Coin Show News Team·October 2, 2026·4 min read

The Ethereum Foundation says its new zkAPI system lets users pay for AI model access and metered APIs without ever linking that payment to their identity or prompt history. Built with the Open Anonymity Project, the tool matters to institutional investors because it ties Ethereum’s core value proposition, privacy-preserving settlement, directly to the fastest-growing category of machine-to-machine commerce.

  • zkAPI was developed jointly by the Ethereum Foundation and the Open Anonymity Project.
  • Users deposit ETH or USDC once into a vault contract, then authorize spending with zero-knowledge proofs instead of API keys.
  • Ethereum targets a gas limit above 100 million in 2026, up from 60 million reached in 2025 after the Pectra and Fusaka upgrades.
  • 100M+ 2026 gas limit target versus 60M reached in 2025
  • 8x theoretical blob capacity gain delivered by Fusaka’s PeerDAS
  • 2029 target year for full quantum-resistant Ethereum execution and consensus layers

The Ethereum Foundation said in a blog post published Thursday (October 1) that zkAPI separates payment from identity for AI inference and other metered services. Today, it wrote, “your API key points to an account, the account to a payment method, and every prompt you send joins the record attached to both,” letting providers reconstruct a persistent profile from billing data alone.

Ethereum Foundation Ships Vault Contract and Zero-Knowledge Proof Client

Under zkAPI, a user makes a single onchain transaction, depositing ETH or USDC into a vault contract on Ethereum. That balance becomes a private cryptographic note, spendable only by the depositor and untraceable to the deposit itself.

To authorize spending, a client running locally on the user’s device generates a zero-knowledge proof. The proof confirms a funded note covers the request without revealing which note, deposit, or person is behind it.

The Ethereum Foundation said one proof can authorize either a single API call or an entire session, with a zkAPI server minting a short-lived, dollar-capped API key that lives only in the device’s memory. Double-spending is blocked by a nullifier, a one-way serial number derived from the note’s secret; a repeat nullifier exposes an attempted double spend without revealing any other information. The system runs on Groth16 proofs over the BN254 curve, Poseidon hashing, and a Merkle tree 32 levels deep, with the vault contract verifying proofs onchain at deposit, close, and withdrawal so users can exit even if every zkAPI server disappears. Default local endpoints for OpenAI and Ollama mean existing chat clients and editors can plug in via localhost without redesign, a pattern similar to how AI agent payment tools like Daski are trying to let software pay for services without a persistent account.

Ken Liu, a Stanford computer science PhD candidate working on the Open Anonymity Project, described the design’s lineage within the project’s broader inference work.

It’s like a generalization of OA unlinkable inference that also abstracts payments away.

Ken Liu, Open Anonymity Project

zkAPI Hides Payment, Not Your IP Address or Prompt Content

The Ethereum Foundation’s own documentation is explicit about what zkAPI does not solve. A provider still sees the content of every prompt and the requester’s network metadata, including IP address, and can attempt to correlate sessions by timing even when billing is anonymous.

The foundation recommends routing requests through Tor with a fresh circuit per session for stronger network anonymity. On the content side, reused personal details, writing style, or project documents can re-link sessions as a fingerprint, a tradeoff the foundation calls inherent to balancing privacy against useful context.

One mitigation floated in the documentation is using local or trusted-execution-environment models to generate requests from shared memory rather than having users retype context by hand, an approach that echoes how projects such as XYO’s work with Autonomys try to verify machine-generated data without exposing its origin. For institutional compliance teams, the open question is whether regulators treat a privacy-shielded billing layer differently from the unlinkable inference layer it sits on top of, since zkAPI explicitly separates the two.

Glamsterdam Upgrade Lands in October as Gas Limit Push Continues

The Ethereum Foundation’s 2026 protocol priorities update, published February 18, organizes work into three tracks: Scale, led by Ansgar Dietrichs, Marius van der Wijden, and Raúl Kripalani; Improve UX, led by Barnabé and Matt; and Harden the L1, led by Fredrik, Pari, and Thomas. The gas limit rose from 30 million to 60 million in 2025 through the Pectra upgrade in May, which added EIP-7702 account abstraction and raised the maximum validator balance to 2,048 ETH, and the Fusaka upgrade in December, which brought PeerDAS and an 8x increase in theoretical blob capacity.

For 2026, the foundation’s roadmap targets a gas limit above 100 million, supported by Block-level Access Lists under EIP-7928. Ethereum educator Anthony Sassano has said 180 million is the year’s ideal outcome, though not the guaranteed one.

October is set to bring Glamsterdam, a dual-layer upgrade pairing the Amsterdam update on the execution layer with Gloas on the consensus layer, widely described as the network’s biggest step since the Merge in 2022. Amsterdam introduces parallel transaction processing on the execution layer, while the Harden-L1 track has already begun implementing quantum-resistant cryptographic algorithms, with full quantum resistance targeted for 2029, a parallel effort to initiatives such as the post-quantum upgrade proposed for Lightning’s offchain layers.

The CCS read. zkAPI is less a product launch than an infrastructure bet: if AI agents start transacting at scale, the entity that owns the anonymous settlement layer owns a toll booth on machine commerce. For ETH holders, that reframes gas demand from speculative trading to recurring, metered inference payments, a stickier revenue base than most L2 fee flows institutions currently model.

The Open Anonymity Project’s GitHub repository and the ZkApiVault contract are already live on mainnet for developers to test, while the harder test comes in October when Glamsterdam ships alongside the push past 100 million gas, the point at which network throughput and zkAPI’s proof-generation costs will be measured under real load.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe