Vitalik Buterin sets security as top priority for zero-knowledge Ethereum

What happened

Vitalik Buterin says the race for speed in the zero-knowledge Ethereum era is over, and security should now come first. In a post on X, he congratulated the ZKsync Airbender team for a new milestone and stated that the fight for speed is definitively won. He wants teams building zkEVMs, which are tools that prove Ethereum transactions with math, to focus fully on security from here.

He named three goals for formal verification. The first is proving that the mathematics behind the proofs matches the Ethereum spec exactly. The second is a hard maximum on prover cost for the amount of gas in a block. The third is the security of the proof system itself. Ethereum's roadmap leans more on zero-knowledge proofs to check blocks, so a proof that is fast but incorrect is dangerous.

Separately, Buterin revealed a privacy-focused AI agent stack. In an X post on October 16, he integrated zkAPI, Tor, random cover traffic, and local LLM prompt rewriting into the pi agent harness. Each AI conversation uses a separate Tor circuit, with random cover traffic added to make requests harder to track. Payment privacy is handled through zkAPI, which Buterin said prevents users from exposing their identities through payment information when accessing AI services.

The setup runs through an OpenAI compatible interface and uses sandbox restrictions to route network activity through Tor. Buterin has published the relevant code on GitHub, although he acknowledged that the system introduces substantial delays and remains vulnerable to certain forms of traffic analysis. He said the setup was becoming usable but acknowledged slower response times and weaknesses against coordinated traffic analysis.

Why it matters

The security message sets a direction for Ethereum's technical road. If teams follow it, the network could end up with proofs that are fast, cheap, and trusted, which is what the roadmap needs. A prover that costs too much for heavy blocks may also push smaller operators out, which hurts decentralization. Vitalik is calling on developers to close those gaps before the technology carries real weight.

The privacy work matters because integrating AI assistants into the Web3 ecosystem has exacerbated the privacy problem: by sending queries to third-party cloud models, users expose their metadata and digital footprint. Buterin's implementation brings Tor routing, zkAPI payments, cover traffic and local prompt rewriting together inside a daemon that supports the OpenAI API interface, allowing the agent to send requests through the privacy wrapper.

The technology's exit from the experimental stage depends directly on the economics of the blockchain itself. Currently, integrating zkAPI into mainnet requires high costs for ZK transactions. The Ethereum update announced by Buterin to reduce gas costs for ZK operations should solve this problem, creating a financially advantageous environment for the operation of autonomous and confidential AI assistants. He expects cheaper zkAPI deposit and withdrawal costs on Ethereum mainnet, although no deployment date has been confirmed.

What is still unclear

• Buterin did not claim that rewritten prompts eliminate every possibility of identifying a user through the information they submit.
• He acknowledged that classic Tor is far from an optimal solution and is vulnerable against active threats: if an AI model provider starts manipulating traffic timings in collusion with the network's entry node, anonymity will be compromised.
• Cover traffic offers reasonable protection against passive observers, but it is weaker if an LLM provider colludes with a Tor guard node and actively manipulates timing.
• In Buterin's opinion, real security for the industry requires a transition to specialized mixnets, where one will have to accept the "tradeoff of a higher bandwidth multiplier" and built-in constant background traffic.

Questions readers ask