Vitalik: AI agents require ZK cryptographic privacy to prevent API calls from exposing user behavior

Ethereum co-founder Vitalik Buterin calls for the development of cryptographic privacy mechanisms to protect API access security and privacy payments in the AI agent era using ZK technology.
(Background: Vitalik casually mentions the “Ethereum One-Year Leap Plan”: after L1 scaling, throughput will increase tenfold)
(Additional context: The Ethereum Foundation reorganizes its R&D team into “Protocol,” focusing on expanding L1 and Blobs, and improving user UX experience)

When you let AI help you plan your schedule, search for medical options, or manage assets, have you considered that your “behavior patterns and thoughts” are continuously leaking to service providers through API calls? Ethereum co-founder Vitalik Buterin warned today (8th) on X: without cryptographic privacy and anonymous payment mechanisms, AI could become the end of human privacy.

Vitalik pointed out that even if AI runs locally, as long as it needs to call external services (APIs), the user’s “search path” is fully visible. Traditional HTTP API calls using access keys or tokens allow service providers to easily infer user intent.

He emphasized that behavioral data has greater inferential power than raw data; service providers can deduce users’ health, financial, and political tendencies from API access patterns.

Crypto privacy is needed if you want to make API calls without compromising the information of your access patterns.

e.g., even with a local AI agent, you can learn a lot about what someone is doing if you see all of their search engine calls

first-order solution to that is to…

— vitalik.eth (@VitalikButerin) March 8, 2026

ZK Technology and Mixnets: Blockchain’s Privacy Solutions

To address these issues, Vitalik proposed a multi-layered technical approach. First is “Mixnets,” which use network-level hops to prevent service providers from knowing who made the request, thus hiding the request source.

Second is “ZK API Payments,” utilizing zero-knowledge proofs (ZK) to enable payments without revealing identity, along with encrypted reputation mechanisms to prevent malicious abuse, such as DoS attacks.

Compared to traditional centralized verification via Azure or OpenAI, the encrypted approach advocates running large language models (LLMs) locally, combined with TEE (Trusted Execution Environment) and on-chain identity standards like ERC-8004, allowing models to run on the user’s device and generate encrypted proofs, ensuring that computation remains confidential.

Ethereum: Privacy Infrastructure for the AI Era

In Vitalik’s vision, Ethereum’s role has evolved beyond mere financial assets to become the privacy infrastructure needed in the AI age. As AI becomes an extension of human capability, blockchain technology is poised to be a key tool to ensure this power is not abused by centralized entities.