OpenAI, Muse and the Agent Economy: Who Gets to Join the Market?

OpenAI gave ChatGPT a wallet. But who gets to decide who can use it?

Your AI being able to buy things isn’t the interesting part.

It’s who gets to authorize the transaction.

OpenAI’s new Assistant Wallet lets ChatGPT agents shop using Visa, Mastercard, and American Express. You’ll still need to approve purchases with your fingerprint or face, and the agent never sees your actual card number. Currently available to US Pro subscribers.

Good design. I’d use it.

But let’s not pretend OpenAI is breaking entirely new ground here.

Meta’s Muse is already doing much of the same thing. It can find products, compare prices, make purchases, track deals, and even handle returns. Purchases require your approval, and Muse uses one-time card numbers through its payment infrastructure to protect your actual card details.

So the question isn’t simply whether AI agents can shop. They can. And increasingly, they can do a lot more than shop.

The more interesting story is what’s happening around them.

The agent economy is quietly developing a guest list.

Payment networks want to identify legitimate agent transactions. Merchants are deciding which bots they will allow through their doors. Airlines and retailers are restricting agents they don’t recognize or authorize.

And that changes the competitive landscape.

The next competitive advantage may not be building the smartest agent. It may be controlling the infrastructure that lets the agent operate.

Think about it.

The big AI companies have the resources, partnerships, payment relationships, and distribution to build agents that work within these approved ecosystems.

What about the independent developer building something genuinely useful in a garage?

Does that agent get the same access? The same payment capabilities? The same opportunity to compete?

Or does it need permission from the very companies it hopes to compete against?

This is where I think things get interesting.

As agentic commerce grows, payment networks, platforms, merchants, and eventually regulators will have an increasingly important role in defining what constitutes a legitimate agent.

There are legitimate security and fraud concerns. Nobody wants rogue agents making unauthorized purchases or abusing merchant systems.

But security and competitive access aren’t necessarily the same thing.

And we should be careful not to build an ecosystem where today’s biggest players get to define the rules for everyone else.

None of this diminishes the technology. Agents that take tedious tasks off our plates are a genuinely exciting development.

I just want the fight over who gets to build them and who gets to use them to remain a fair one.

So here’s the question: When your AI agent goes shopping, who decides whether it’s legitimate? And why should the companies that profit from the agent economy also get to decide who participates in it?

Leave a Reply