Recently, Google released the Agent Payments Protocol (AP2), a payment protocol designed specifically for AI Agents.

When I saw this news, my first reaction was that the Internet giant has finally extended its tentacles into the payment field of AI Agents.

Why do I say "finally"?

Because AI Agents payment is not unfamiliar to users who have been paying attention to the crypto ecosystem in the past two years, the AI Agent BYTE in the Virtual ecosystem has long implemented payment applications and has been integrated into Coinbase's application.

In the payment scenario implemented by BYTE, it uses the crypto asset USDC for payment.

This formed my original impression of AI Agents payment: this scenario is naturally suitable for and should use crypto assets.

Therefore, I have always believed that the advancement of this field should be driven by native companies in the crypto ecosystem or at least companies with crypto businesses.

In addition, when using crypto assets for payment, all transactions are on-chain, which inherently possesses the most important core characteristics of payment services such as transparency, verifiability, and non-tamperability.

Since the crypto ecosystem already has these features, and traditional Internet giants do not have them, in this case, what else can Google do besides developing its own L1 for its entry into the financial business?

With this curiosity, I studied its announcement content (see reference link for details).

I think the two typical scenarios listed in this announcement best reflect the purpose of publishing this protocol:

- Live shopping (with a human present). A human instructs the AI Agent to perform a purchase and ultimately signs to approve the payment. During this process, the human's instructions are recorded to prove that the tasks performed by the AI Agent were indeed issued by a human and not a fraudulent act.

- Proxy tasks (human absence). A human pre-arranges a purchase to be executed by an AI agent in their absence. This pre-arranged instruction is recorded to prove that the task executed by the AI agent was indeed pre-arranged by the human, that it was executed due to certain conditions being triggered, and that it was not fraudulent.

From the content of this announcement, I understand that the core of Google's AP2 protocol is not to reshape the process of on-chain payment transactions, but to prove that the payment transactions executed by AI agents have definite reasons and sources, more precisely, from humans.

Google’s announcement listed several well-known Ethereum ecosystem partners, such as the Ethereum Foundation, EigenCloud (formerly EigenLayer), and Coinbase.

The Ethereum Foundation recently established a new organization dedicated to AI applications, specifically verification within them. This demonstrates that the Ethereum core team is taking the AI race seriously and doesn't want to be left out of this crucial race for the future. I wonder if this move is directly related to Google's agreement?

The original intention of establishing EigenCloud/EigenLayer was to leverage the security of the Ethereum network to provide security services for other projects that also needed it. However, several years later, aside from the initial buzz surrounding the token launch, the project has remained largely silent, let alone any projects achieving significant success using its services. We hope that AI Agents Payment will bring real applications and scenarios to this project and its related sectors.

I think Coinbase is more likely to directly benefit from this protocol.

The reason is simple, because Coinbase's current application integrates several AI Agents that have already implemented payment applications. If this protocol is supported, its encrypted payment closed loop will be more perfect and secure.

On-chain payment was originally a scenario envisioned by AI + Crypto in the crypto ecosystem. Now traditional Internet giants are also using this scenario as an attempt to enter the AI + Crypto track.

The results of these two ecosystems exploring from different angles surprisingly reach the same conclusion, which may indicate that in the near future we will really be able to see disruptive applications emerge in this scenario.

Reference link: https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol