#以太坊三季度涨70.9%
From the surface, it looks like Ethereum has launched yet another privacy tool. But what really needs to change might be how AI gets paid..
📢 最新消息群里说
On Thursday, the Ethereum Foundation released a system called zkAPI, developed together with the Open Anonymity Project, and it’s already running on the Ethereum mainnet. The usage is not complicated: you deposit ETH or USDC into a vault contract, and the balance is recorded as a private ticket. Then, each time you call an AI interface, your device locally computes a zero-knowledge proof to show that the funds are sufficient—without telling the service provider which specific ticket it is. As a result, the service provider and the payment server can’t link the request to the payer. Every payment also includes a sequence number called a nullifier. If someone tries to spend the same balance twice, it will collide and fail. What gets exposed is only this..
Most people see it as another privacy project. But the real goal here isn’t privacy—it’s machine-to-machine payments. Right now, every AI API call is tied to an identity: your key points to an account, and the account is linked to a payment method. Every prompt you’ve ever sent is attached to that same record, and the service provider can stitch together years of usage into a person’s profile. If in the future AI agents handle everything for you—calling models, buying compute, and paying for bandwidth—payments between machines shouldn’t be attached to any person’s ID. zkAPI fills exactly this gap: payments can be proven, but identity doesn’t have to be disclosed..
The timeline is worth taking a closer look.. The Ethereum Foundation’s dAI team was established in September 2025, aiming to make Ethereum the settlement and coordination layer for AI. Earlier this year, in January, they put their AI agent identity standard, ERC-8004, on the mainnet. The zkAPI whitepaper was jointly released this February by Vitalik and dAI lead Davide Crapis. In other words, the thread—identity, payments, settlement—has been drawn out for a year, not something they rushed in just to ride the AI wave..
It’s also interesting on the money side.. This privacy segment has been gaining momentum throughout the year, but the market mostly treats it as speculation. What Ethereum wants to do is different: it aims to turn privacy from a tool for hiding funds into AI infrastructure. If that direction takes hold, the pricing logic for privacy assets would shift from pure hype to real usage. And once usage rises, ETH—serving as that settlement layer—would gain yet another push..
Of course, don’t rush to hoist the banner yet.. The official documentation is very clear: zkAPI does not provide anonymity at the network layer. The gateway may still associate requests via a stable IP, and sessions may also be re-linked based on the contents of the prompt—personal information, writing style, chat history. It’s more like “payments without a trace,” not “no one knows when you use it.” The repository itself also labels it experimental..
So what’s really worth watching isn’t its launch, but two questions.. First, will there be real AI usage flowing through this channel, or will it just be a small group of geeks entertaining themselves? Second, will regulators treat an unidentifiable payer as a direct compliance issue and investigate it? The first determines whether this becomes infrastructure or a toy. The second determines how long it can survive. As long as either one doesn’t pan out, the narrative of AI plus privacy is still just a narrative..
From the surface, it looks like Ethereum has launched yet another privacy tool. But what really needs to change might be how AI gets paid..
📢 最新消息群里说
On Thursday, the Ethereum Foundation released a system called zkAPI, developed together with the Open Anonymity Project, and it’s already running on the Ethereum mainnet. The usage is not complicated: you deposit ETH or USDC into a vault contract, and the balance is recorded as a private ticket. Then, each time you call an AI interface, your device locally computes a zero-knowledge proof to show that the funds are sufficient—without telling the service provider which specific ticket it is. As a result, the service provider and the payment server can’t link the request to the payer. Every payment also includes a sequence number called a nullifier. If someone tries to spend the same balance twice, it will collide and fail. What gets exposed is only this..
Most people see it as another privacy project. But the real goal here isn’t privacy—it’s machine-to-machine payments. Right now, every AI API call is tied to an identity: your key points to an account, and the account is linked to a payment method. Every prompt you’ve ever sent is attached to that same record, and the service provider can stitch together years of usage into a person’s profile. If in the future AI agents handle everything for you—calling models, buying compute, and paying for bandwidth—payments between machines shouldn’t be attached to any person’s ID. zkAPI fills exactly this gap: payments can be proven, but identity doesn’t have to be disclosed..
The timeline is worth taking a closer look.. The Ethereum Foundation’s dAI team was established in September 2025, aiming to make Ethereum the settlement and coordination layer for AI. Earlier this year, in January, they put their AI agent identity standard, ERC-8004, on the mainnet. The zkAPI whitepaper was jointly released this February by Vitalik and dAI lead Davide Crapis. In other words, the thread—identity, payments, settlement—has been drawn out for a year, not something they rushed in just to ride the AI wave..
It’s also interesting on the money side.. This privacy segment has been gaining momentum throughout the year, but the market mostly treats it as speculation. What Ethereum wants to do is different: it aims to turn privacy from a tool for hiding funds into AI infrastructure. If that direction takes hold, the pricing logic for privacy assets would shift from pure hype to real usage. And once usage rises, ETH—serving as that settlement layer—would gain yet another push..
Of course, don’t rush to hoist the banner yet.. The official documentation is very clear: zkAPI does not provide anonymity at the network layer. The gateway may still associate requests via a stable IP, and sessions may also be re-linked based on the contents of the prompt—personal information, writing style, chat history. It’s more like “payments without a trace,” not “no one knows when you use it.” The repository itself also labels it experimental..
So what’s really worth watching isn’t its launch, but two questions.. First, will there be real AI usage flowing through this channel, or will it just be a small group of geeks entertaining themselves? Second, will regulators treat an unidentifiable payer as a direct compliance issue and investigate it? The first determines whether this becomes infrastructure or a toy. The second determines how long it can survive. As long as either one doesn’t pan out, the narrative of AI plus privacy is still just a narrative..
