The Agent Fee Is the Prize, Blocking Is the Defense — and JPMorgan Chase's New Terms Send the Fraud Bill to Shared Passwords

OpenAI charges Shopify merchants 4% on sales made through ChatGPT checkout, on top of Shopify's own platform and payments fees, effective Jan. 26, 2026. Meta co-wrote an open agent standard with Walmart, Shopify and Stripe on Oct. 6, 2026, but left payments as a future extension. Amazon and JPMorgan Chase take nothing on an agent order because they refuse it. Every platform in this market is choosing between a fee and a wall.

The sequencing matters. Amazon blocked first, on Sept. 20, 2026. On Oct. 6, 2026 the open standard was signed by Meta, Walmart, Shopify, Stripe, Genesys, Instinct and Rocket — every interested party except the blockers. JPMorgan Chase closed the write path a week later, on Oct. 13, 2026, and Walmart opened its gateway eight days after that. The standard was written by the platforms the blockers exclude, and the payment layer that decides who gets paid was deliberately left unspecified.

China skipped the argument entirely. Alibaba connected its Qwen assistant to the full Taobao catalog and launched a Qwen-powered shopping assistant inside the Taobao app in May 2026. The U.S. is arguing about whether to admit agents at the same moment one of the largest retail platforms on earth already treats the agent as the store.

A benchmark released Oct. 29, 2026 and reported Nov. 1, 2026 ran six widely used bot-detection products against deliberately disguised agent sessions built with stealth browsers and session takeover. BotD caught 14.8% of them, Akamai 8.7%, F5 6.8%, Cloudflare Turnstile 4.6%, Alphabet's reCAPTCHA 3.6% and Vercel BotID 3.1%. No mainstream product identified even one in five disguised agents, which is why Amazon and the banks refuse agents outright instead of trying to authenticate them.

That is why the blockers' language sounds like regulation and their behaviour looks like pricing. The Federal Reserve has named trust between buyers and sellers as the principal barrier to scaling agentic commerce, and the standard answer to a trust problem is an agreement about who signs, who verifies and who pays. The agreement is being written now, and the platforms that refused to join it are the ones that will have to accept someone else's rules.

The banks did not simply close the door. They rewrote the terms for anyone who had already handed over a password. Changes to the Chase Digital Services Agreement, effective Oct. 27, 2026, state that third parties including External AI are prohibited from using a customer's login credentials, that a customer who permits such access is deemed to have authorized the agent's transactions, that the customer is responsible for errors including transactions the agent completes beyond the authority given, and that the customer must indemnify the bank and monitor the agent themselves. The customer now carries the loss on transactions the bank never reviews.

The legal ground shifted under the blockers at the same time. Amazon won a preliminary injunction against Perplexity in March 2026 and lost it on Aug. 4, 2026, when the Ninth Circuit held that the user, not the AI company, was accessing Amazon's computers. Amazon stopped calling agents hackers and started calling them a breach of its Conditions of Use — and the bank did the same thing by burying the risk in a digital services agreement. Contract, not statute, is becoming the load-bearing instrument of agentic commerce.

Amazon said it cut off Meta's Muse because the agent did not identify itself and appeared to capture and store customer credentials. That is the stated reason. The measured reason sits in the segment data. Advertising services revenue in the third quarter of 2026 was $19.64 billion, up 10.9% from $17.70 billion a year earlier, while total net sales rose 12.1% to $201.94 billion. Advertising grew more slowly than the rest of the business — in the same quarter a year earlier it grew 23.6%.

Shopify is the cleanest beneficiary of the split because it monetizes both sides of the transaction. In the quarter ended June 30, 2026, GMV reached $115.6 billion, up 32%, and 68% of it ran through Shopify Payments, up from 64% a year earlier. For the first six months, merchant solutions revenue of $5.2 billion grew 38% against 22% for subscription solutions. Payments are growing nearly twice as fast as subscriptions, and every agent-driven order routes through them.

That is the asymmetry investors should notice. Both camps pay the same security and payment suppliers, and a blocked agent still settles on the same card network, which is why Mastercard's net revenue grew 35% in the very quarter Amazon was cutting agents off and JPMorgan Chase was closing accounts. The upstream layer does not need to pick a winner.

That59% is the most uncomfortable number for the blockers. Merchants priced the agent fee within weeks while consumers inherited the fraud loss through a contract amendment — the two sides of the same market are being repriced on completely different clocks.

Over the next two quarters the measurable question is volume: Walmart's November agent launch and the holiday peak are the first clean test of whether agent sessions buy or merely browse. Over one to three years the decisive question is whether the write path ever reopens at banks, because the day it does the agent becomes the account holder and the merchant relationship moves to whoever holds the agent's consent. Neither camp wins outright: the platforms that open gates keep the fee and carry the fraud cost, and the platforms that close them keep the customer and the ad money but cede the transaction to whoever arrives with an approved protocol.

Two things to watch. First, who sets the price: the standard that Meta, Walmart, Shopify and Stripe signed still lists payments as a future extension, so the fee on an agent transaction is undefined and the first party to write it down captures the economics. Second, what Shopify prints on Nov. 3, 2026: a third quarter in which agentic volume appears in merchant solutions would confirm that the fee-taker camp is compounding while the blockers are still arguing about whose liability it is.