Vitalik Buterin and Ethereum Foundation AI Lead Davide Crapis have proposed a new privacy-preserving payment framework designed to fundamentally change how APIs are accessed in the age of AI. In a newly published research post titled “ZK API Usage Credits: LLMs and Beyond,” the duo argue that the internet’s identity-based access model is increasingly incompatible with AI-driven systems.

Their proposal replaces identity with stake, using zero-knowledge cryptography to enable anonymous, secure, and efficient API usage at scale. Instead of authenticating every API request with an email address, credit card, or persistent wallet identity, the proposal suggests a different model.

It is a shift from identity-based enforcement to stake-based cryptographic accountability. And in the age of autonomous AI systems, that shift could be critical.

• Why Identity-Based Access Is Becoming a Liability
• The Structural Limits of Web2 Billing & On-Chain Payments
• Understanding the ZK API Credits Model
• Ethereum’s Expanding AI & ZK Strategy

Why Identity-Based Access Is Becoming a Liability

For decades, internet access has meant identity. To use a service, you create an account.

You verify an email. You attach a payment method. Every API request is linked to a persistent profile. The system works because identity enforces accountability.

But AI fundamentally changes the scale and sensitivity of API usage. Large language models process confidential legal drafts, health data, enterprise research, internal communications, and financial strategy documents.

At the same time, AI systems can generate thousands of API calls automatically. When each of those calls is tied to a persistent identity, the metadata itself becomes revealing.

Even if content remains encrypted, behavioral patterns do not. This concern becomes more urgent as AI systems begin transacting with each other. EtherWorld recently covered the first AI-to-AI commercial transaction going live on Polygon, marking a symbolic step toward machine-driven commerce.

The Structural Limits of Web2 Billing & On-Chain Payments

The Web2 billing model ensures payment through identity verification. Email logins and credit cards create clear accountability. But they also create central points of data aggregation and profiling.

The internet made access identity-based.

Every API call tied to your email, card, or wallet.

AI will make it worse.

Our proposal:
Replace identity with stake.

Deposit once.
Make thousands of API calls.
Stay unlinkable.
Abuse gets slashed.

New post with @VitalikButerin 👇 pic.twitter.com/xUAA2cCUtJ— Davide Crapis (@DavideCrapis) February 11, 2026

For enterprises experimenting with AI, this can be a major barrier. Sensitive prompts tied to identifiable accounts introduce legal and operational risks.

The alternative in Web3 is per-request on-chain payment. This model guarantees payment cryptographically but introduces inefficiencies. Transactions cost gas. They require confirmation time. And most importantly, they expose usage patterns publicly.

Even as Ethereum explores zero-knowledge rollups and scaling upgrades, linking frequent API calls directly to on-chain transactions is not practical for high-throughput AI workloads. This tension reflects a broader theme in Ethereum’s evolution.

As previously explored in EtherWorld’s coverage of ERC-8004 and Ethereum’s move toward agentic AI markets, the protocol is gradually preparing for a world where autonomous agents transact, coordinate, and consume services programmatically. That future requires infrastructure that is efficient, private, and economically secure.