American Express Introduces Framework to Prevent Unauthorized AI Purchases

American Express has introduced a new framework designed to prevent AI agents from making unauthorized purchases on behalf of their users.

The growing use of personal AI agents for transactions raises concerns about the potential for AI "misfiring" and initiating unauthorized payments due to the unpredictable nature of large language models.

To mitigate this risk, Amex's Intent Intelligence Framework aims to capture and verify a customer's purchase intent before an AI agent completes a transaction. This ensures that the customer's instructions are consistently carried through from start to finish.

The framework is part of Amex's Agentic Commerce Experiences (ACE) Developer Toolkit, which was launched in April and includes several services designed to facilitate agentic transactions.

Amex is taking a proactive approach to agentic commerce by building technology that prevents AI agents from making unauthorized purchases, despite not yet seeing an increase in disputes related to this type of commerce.

A recent survey conducted by Amex found that 66% of U.S. consumers are likely to use AI agents to research the best price for a product or service within the next 12 months, while 61% said they would use it for recommendations during the same period.

The new intent framework developed by Amex begins with a Create Intent API, which captures customers' purchase instructions and other constraints associated with their requests. This flexibility is designed to accommodate various AI models that may be used in agentic commerce.

To ensure secure transactions, customers must authenticate their identity and sign off on the intent captured by the agent. This process generates an Intent ID and a Proof of Intent Token, which includes high-level customer constraints.

The Proof of Intent Token is then used to obtain a digital payment credential, allowing agents to complete purchases while keeping customers' credit card numbers secure.

Amex's Intent ID and token can be used throughout various stages of a transaction, enabling companies to effectively manage risks related to fraud prevention and dispute resolution. This approach is particularly useful in preventing unauthorized transactions by flagging discrepancies between the customer's intent and the actual purchase.

According to Amex, its framework differs from others in that it captures intent earlier on in the transaction process, thereby reducing risk before it occurs. For instance, if a customer instructs an agent to buy white Nike sneakers in size 9 but the agent mistakenly submits a request for Adidas shoes, the transaction can be blocked due to the discrepancy.

Amex foresees a future where agents may increasingly operate autonomously, acting on behalf of customers without explicit consent in certain situations. In such cases, technology will play a crucial role in verifying that transactions align with the customer's original intent. This is particularly relevant for low-stakes purchases like re-ordering household essentials.

Amex envisions scenarios where agents can automatically complete routine tasks, such as replenishing paper towels or toilet paper, without requiring explicit consent from customers. In these situations, technology will be essential in ensuring that transactions accurately reflect the customer's original intent and are therefore legitimate.

A log of authorized transactions is crucial for addressing potential agent-related fraud, according to industry experts.

This log would track what was approved by each AI agent and when, helping to prevent unauthorized purchases made on behalf of the agent.

Amex is already in talks with various partners across the payments ecosystem to implement its ACE framework, which includes agentic tokens containing Amex's payment credentials.

As part of this development, Amex plans to continually update and enhance the ACE developer kit as the agentic commerce landscape evolves.

Amex has developed an agentic commerce (ACE) framework to prevent artificial intelligence agents from making unauthorized purchases. This framework is designed to ensure that AI-powered shopping assistants operate within predetermined boundaries and do not exceed their intended spending limits.

The ACE developer kit will allow businesses to integrate the framework into their systems, enabling them to control how AI agents interact with their e-commerce platforms. As part of this development, Amex plans to continually update and enhance the ACE developer kit as the agentic commerce landscape evolves. This ongoing process aims to provide a secure and reliable solution for companies seeking to leverage AI in their shopping experiences.