“Target is not responsible for the conduct of the third-party Agentic Commerce Agent.”

That sentence is now in Target’s official terms of service. They added it on March 22. No press release. No announcement. Futurism picked it up April 5.

Here’s the plain English version. When an AI shopping agent buys something on your behalf at Target, that purchase is treated as if you made it yourself. If the agent buys the wrong size, the wrong color, the wrong product entirely, that’s on you. If the agent gets confused and orders five of something you wanted one of, that’s on you. If the agent gets tricked or hacked and buys something fraudulent, that’s on you.

The exact ToS language:

“Actions taken by an Agentic Commerce Agent within the scope of your approved permissions will be treated as actions authorized by you. You are responsible for reviewing activity performed by your Agentic Commerce Agent and for promptly notifying the Agentic Commerce Agent and Target of any activity you believe is unauthorized or outside the scope of permissions you approved.”

“Target does not purport to guarantee that an Agentic Commerce Agent will act exactly as you intend in all circumstances.”

The only consumer protection a Target spokesperson offered was that products bought via agents are still eligible for returns through Target’s standard return policy. Still eligible — not guaranteed.

Walmart did the same thing. Their ToS for Sparky now warns that generative AI “may not be accurate, complete or up-to-date and may be misleading.” Same liability shift, same logic: the user authorized the agent, so the user owns the outcome.

Amazon hasn’t done this yet. That’s worth watching. They either follow Target and Walmart in the next quarter, or they position themselves as the only major retailer that stands behind agent-initiated purchases. That’s a marketing wedge that writes itself.

Every conversation about agentic commerce until now has been about whether the technology works. Can the agent find the right product? Can it complete checkout? Can the payment rails handle it? Can the catalog data answer the question correctly?

The bigger question is the one nobody was asking: who pays when it doesn’t?

Target just answered. You do.

This is the legal layer of agentic commerce being written in real time, buried in terms-of-service updates consumers will never read, before anyone has tested whether the language survives a class action. Juniper Research projects $8 billion in agent-initiated transactions for 2026 alone. By the time the first lawsuit hits, serious money will have already gone through under these terms.

There’s a real reason Target moved first. They co-built the Universal Commerce Protocol with Google. They have a ChatGPT beta app live. They made agentic commerce a major strategic focus at NRF in January. They’re already running these transactions at volume. The ToS update isn’t theoretical. It’s defensive.

I’ve been telling brands at every conference this year that agentic commerce is moving faster than the industry thinks. Target just made it more real than I was prepared for.

If you sell on Target, this should bother you. When an agent buys the wrong variant of your product and the consumer eats the loss, that’s not actually contained to the consumer. It comes back to you. Bad review. One-star rating. Refund demand. Support ticket. Returned item. The customer didn’t pick the wrong size. The agent did. But you’re the one who hears about it. And you have no relationship with the agent that made the mistake. You don’t know which model, which prompt, which context window produced the error. You just know you have an angry customer who blames your product.

The retailer just opted out of the relationship. The brand can’t.

If you’re a consumer, the worry is different. You’re being asked to trust agents with your money before the legal protections exist. The pitch is “let the agent handle it.” The reality is “let the agent handle it, and if it screws up, that’s your problem.” Those are different products.

So what should you actually do with this?

  1. If you sell on Target, start tracking agent-initiated returns separately from consumer returns. They’ll look identical in your data at first. The cause is different and the trend will be different. If you can’t tell them apart, you can’t price for them.
  2. Push your retailer reps for an answer on the brand side. Ask Target, Walmart, and Amazon directly: when an agent buys my product wrong, what’s your policy on chargebacks, returns, and reviews? Most of them won’t have an answer yet. Force them to write one.
  3. Watch Amazon. If they decline to follow Target and Walmart, that’s a competitive position they can build on. If they follow, the entire major-retailer category has formally said “we don’t stand behind agent purchases.” That changes how brands should think about which channels to feed.
  4. Factor agent error into your return rate forecasts on Target now. Even if it’s small at first, the policy is in place. The volume will grow with self-serve agent access opening later this year.
  5. If you’re a consumer, read the terms before you authorize an agent to shop on your behalf. You’re agreeing to wear the loss when the tool gets it wrong. That’s the deal you’re signing.

The first rule of agentic commerce just got written. It says the consumer pays. Every other rule will get built on top of that one.