American Express Develops New Framework to Prevent Unauthorized Purchases by AI Agents
American Express has introduced a new system called the Intent Intelligence Framework to prevent errors by large language models and secure customer authorization.
American Express has announced a new security system called the Intent Intelligence Framework to prevent potential errors and unauthorized spending during transactions carried out by AI agents on behalf of users.
New Security Framework for AI Agents
As personal AI agents increasingly perform transactions on behalf of users, the unpredictable nature of large language models can lead bots to take unintended actions.
To prevent this issue, American Express has developed a specialized framework to capture and verify customer intent before the AI agent completes the transaction.
Operating Principle and APIs of the System
The new framework initiates the process with the Create Intent API, which records the customer's purchase instructions, merchant name, maximum budget, and product preferences.
After verifying the customer's identity and confirming the intent, secure digital payment information is used instead of the actual credit card number, thanks to the identifiers and tokens generated.
Discussions with Industry Stakeholders
American Express has integrated this new intent framework into its Agentic Commerce Experiences developer toolkit, which was launched in April.
The company aims for this new system to become the industry standard for AI-driven transactions and is holding discussions with partners in the ecosystem.