American Express has introduced its Intent Intelligence Framework, a new security approach designed to help prevent AI agents from making purchases that customers did not authorize.
Announced on October 13, 2026, the framework aims to ensure that an AI agent follows a customer's original purchasing instructions throughout the transaction process. As AI-powered shopping assistants become more capable of researching products, comparing prices, and completing purchases, payment providers are working to ensure that automation does not compromise customer control.
How the Intent Intelligence Framework Works
The framework captures a customer's purchasing instructions and uses them to help validate transactions before an AI agent completes a payment.
Its process includes four key steps:
Capture purchase intent: An API records the customer's instructions, including product preferences, the preferred merchant, and spending limits.
Authenticate the customer: The customer verifies their identity and approves the captured purchasing instructions.
Create proof of intent: The system generates a unique Intent ID and a Proof of Intent Token representing key purchase constraints.
Validate the transaction: The token can be used to obtain a secure digital payment credential, helping ensure the transaction aligns with the customer's approved instructions.
For example, if a customer authorizes an AI agent to purchase white Nike sneakers in size nine, but the agent attempts to buy Adidas shoes instead, the framework is designed to help identify the mismatch before the transaction proceeds.
Why AI Shopping Needs Additional Security
AI agents are increasingly being developed to complete tasks on behalf of users, including shopping, booking services, and managing recurring purchases. However, AI systems can misinterpret instructions or take actions that do not match a customer's expectations.
Traditional payment security primarily focuses on verifying payment credentials and detecting suspicious transactions. Agentic commerce introduces another question: Was the purchase itself consistent with what the customer authorized?
By making purchase intent available earlier in the transaction lifecycle, American Express aims to add another layer of control. The framework may also help payment providers investigate disputes and identify which agent was authorized to act.
What This Means for the Future of Digital Payments
American Express said it had not yet observed an increase in disputes caused by agentic commerce but was developing safeguards ahead of wider adoption.
The company's approach reflects a broader shift in digital payments, where AI agents may eventually make purchases with limited direct human involvement. In these scenarios, security systems will need to account for the customer's identity, the agent making the purchase, and the boundaries of the customer's authorization.
American Express has also said it intends to support common industry standards so intent-based safeguards can work across different agentic commerce models.
The framework is designed to reduce the risk of unauthorized or incorrect purchases, but it should not be interpreted as a guarantee that every AI-driven transaction will be error-free.
