Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

American Express Introduces a Framework to Stop AI Agents From Making Unauthorized Purchases

American Express has introduced a new framework designed to ensure AI agents only complete purchases their users have authorized. As AI-powered shopping grows, the mechanism aims to head off costly mistakes and payment disputes.

Published
美国运通推出授权框架,防止 AI 代理擅自下单支付
Image source: techcrunch.com

American Express has unveiled a new framework to govern how AI agents make payments. According to TechCrunch, the mechanism's central aim is to guarantee that an AI agent only completes purchases the user has actually authorized, rather than ordering beyond its mandate.

The backdrop is the rise of agentic shopping. Once an agent can compare prices, add items to a cart and even check out on a user's behalf, a single unauthorized transaction can cause real financial loss and the disputes that follow — exactly the problem the framework is meant to get ahead of.

By positioning, American Express frames it as a layer of authorization assurance rather than plain transaction risk control. It tries to draw the line between clear user consent and agent-executed action, so every payment an agent initiates can be traced back to a user's approval.

For a card network, the value of such a framework is that it moves liability assignment earlier. If charges from an overstepping agent can be blocked in time, the cost of resolving disputes across merchants, cardholders and the network falls — and payment certainty actually rises.

The move also signals how payment networks view the commercialization of AI agents: keen to capture the added transaction volume agents bring, yet determined to build guardrails first. Whether merchants and agent developers adopt the framework widely will decide if it becomes an industry default.

What to watch is implementation: how the framework verifies authorization, which transaction scenarios it covers, and whether it interoperates with rival networks' approaches. If each network builds its own authorization standard, agent developers may simply face new integration burdens.

Seen broadly, AI agents are moving from being able to do things toward being able to spend money, and payment authorization is one of the hardest and most critical links in that chain. American Express's attempt offers an early template; how banks, merchants and platforms follow will be worth tracking.

Why it matters

If the framework is adopted widely, AI agent payments would gain a traceable boundary that cuts unauthorized charges and dispute costs, clearing a key obstacle to agent commercialization.

American ExpressAI Agent支付
Back to realtime news

Nearby Updates

All