American Express Unveils Intent Framework to Secure AI Agent Purchases
American Express introduces the Intent Intelligence Framework to capture customer purchase intent and prevent unauthorized transactions by AI agents.
Personal AI agents are increasingly handling transactions on behalf of users, but the nondeterministic nature of large language models means these bots can occasionally misfire. To address the possibility of AI systems making unauthorized purchases, American Express is rolling out a new structure aimed at capturing and authenticating customer purchase intent before a transaction is finalized.
Known as the “Intent Intelligence Framework,” the company hopes the system will evolve into an industry standard for transactions driven by artificial intelligence. Hannah Bozian, VP of Agentic Partnerships and Strategy at Amex, noted in an interview that these new transaction flows involve a consumer, a merchant, and an agent. The goal is to ensure that agentic commerce remains seamless, trusted, and secure.
This release builds upon the April launch of Amex’s Agentic Commerce Experiences, or the ACE Developer Toolkit, which now incorporates the intent framework. While Amex reports it has not yet observed a rise in disputes linked to agentic commerce, the organization is taking a proactive stance as this market expands. According to an Amex Trendex survey, 66% of U.S. consumers indicated they are likely to use AI agents to find the best prices over the next year, and 61% expressed openness to using them for product recommendations.
How the Intent Intelligence Framework Operates
The framework initiates with a Create Intent API, which records the user's purchase instructions and constraints, such as preferred merchants, maximum budgets, and product preferences. Although the method of collecting intent may differ based on the AI model utilized, the architecture is designed to accommodate various agentic commerce setups.
After collection, the customer authenticates their identity and approves the captured intent. This process generates an Intent ID along with a Proof of Intent Token containing the user's constraints. The token secures a digital payment credential that the agent utilizes at checkout, protecting the customer's actual credit card number while ensuring the purchase matches authorized parameters. Additionally, these tokens assist companies with fraud prevention and dispute resolution.
Bozian emphasized that capturing intent early in the transaction flow helps mitigate risks before they materialize. For example, if a user instructs an agent to buy a specific pair of sneakers but the agent attempts to purchase a different brand, the system can block the transaction due to the mismatch. While many current agents require direct consent prior to buying, Amex anticipates a future where agents act more autonomously for routine reorders, making technological oversight essential.
What it means for developers
Developers building autonomous systems must adapt to stricter verification standards as financial institutions introduce guardrails for AI-driven actions. Amex is actively engaging with partners across the payments ecosystem to implement ACE, which includes agentic tokens housing payment credentials. As developers build and test these agentic applications, they can try top AI models cheaply through one API at https://apixoai.online.
Amex plans to continue developing the ACE Developer Toolkit as the agentic commerce landscape evolves. Further details regarding the Intent Intelligence Framework will be discussed by Hannah Bozian at TechCrunch Disrupt 2026.
Source: Amex introduces a framework to prevent AI agents from making unauthorized purchases — TechCrunch AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
