Verifiable intent is a machine-verifiable record that a user authorized an AI agent to take a specific action under specific constraints. It is especially important for payments, purchases, bookings, and other consequential workflows where a service must distinguish an authorized agent action from an accidental or malicious one.
In agentic commerce, verifiable intent can provide evidence such as:
- who authorized the agent;
- which agent or host received the mandate;
- what item, merchant, amount, or constraint was approved;
- when the authorization was issued;
- whether the user was present or absent;
- what payment or checkout evidence was linked; and
- whether the record was tampered with.
Verifiable intent is related to Agent Payments Protocol (AP2), but the concept is broader than payments. Similar evidence patterns may apply to legal approvals, data exports, account changes, enterprise procurement, or privileged tool calls.
The core design goal is accountability. If an autonomous system acts later, the service should be able to verify that the action still matches the user's original authorization and policy constraints.
Google describes Verifiable Intent as a standard for creating a tamper-proof log of user-authorized agent actions in its AP2 donation announcement.
The LLM Knowledge Base is a collection of bite-sized explanations for commonly used terms and abbreviations related to Large Language Models and Generative AI.
It's an educational resource that helps you stay up-to-date with the latest developments in AI research and its applications.