TLDRocket
Sign in

Amex introduces a framework to prevent AI agents from making unauthorized purchases

TechCrunch Sarah Perez

Amex built a new system to check what an AI agent is actually allowed to buy before it pays. It’s meant to stop bots from drifting off-script and making bad purchases.

Based on reporting by TechCrunch, Sarah Perez — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

American Express is trying to put a seatbelt on AI shopping. The company has introduced the Intent Intelligence Framework, a system meant to capture what a customer meant to buy before an AI agent actually completes the transaction. In Amex’s view, that matters because large language models are not exactly famous for sticking to the script.

The framework sits inside Amex’s Agentic Commerce Experiences, or ACE Developer Toolkit, which launched in April. Amex says it hasn’t seen a jump in disputes from agentic commerce yet, but it wants the plumbing in place before this stuff gets bigger. That’s a very payments-company instinct, and probably the right one.

Here’s the basic flow. A Create Intent API collects the customer’s instructions and any constraints they include — things like the merchant’s name, a maximum budget, and product preferences. The customer then authenticates and signs off on that intent. After that, the system creates an Intent ID and a Proof of Intent Token, which carries the high-level constraints. That token is then used to get a secure digital payment credential for checkout instead of the customer’s actual card number.

Amex says the same Intent ID and token can also help with fraud prevention and dispute resolution later in the process. The company’s argument is that intent should be checked earlier, before the wrong purchase ever gets far enough to become a mess. Hannah Bozian, Amex’s VP of Agentic Partnerships and Strategy, gave the example of a customer asking for white Nike sneakers in size 9 and the agent trying to buy Adidas instead. In that case, Amex says, the transaction could be stopped.

The company is also thinking beyond today’s fairly cautious agent setups. Bozian said many current agents ask for consent before buying, but future systems may act more autonomously, especially for routine reorders like paper towels or toilet paper, where the real limits are price and size. Amex says its framework is meant to work across different agentic commerce models, and the company is already talking with partners across the payments ecosystem about ACE as the space develops.

My take — AI-written commentary, not fact-checked reporting

This is the sensible part of agentic commerce: not more hype, more receipts. If an AI is going to spend money, someone should be able to prove what it was told to do, and payers like Amex are right to force that argument early. The industry’s favorite hobby is shipping first and calling it trust later; this at least does the opposite.

Read more about this at: TechCrunch

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.