Only 24% of US adults would feel comfortable letting AI make a purchase for them, a study shows.
Imagine a world where AI can book a flight and order new luggage on your credit card. Don’t spend too long imagining; we’re already living in it.
Visa said this week that it’s partnering with OpenAI to allow you to make Visa payments on its ChatGPT AI chatbot. Visa announced the news at Visa Payments Forum 2026 on Wednesday in San Francisco. Visa plans to use tokenization, which replaces your 16-digit card number with a random string of numbers to safeguard your information when you pay online.
The partnership is a part of Visa Intelligent Commerce, the payment processor’s platform for agentic commerce. The goal is for OpenAI and Visa to join forces to let ChatGPT initiate and complete transactions — without your help. How safe is that?
Creating an entirely new layer to the digital payment landscape can create many problems, says Daniella Flores, an AI and machine learning technical writer, financial and career writer, and former CNET Money Expert Review Board member. “In any situation, not just this one, the more parties that have your payment credentials, the more opportunity there is for data breaches and theft,” Flores said in an email to CNET.
How will ChatGPT work with Visa?
A Visa spokesperson said in an email to CNET that the goal for Visa’s OpenAI partnership with Visa is to make shopping easier and more efficient while making sure consumers still decide when, where and how their money is spent.
“Consumers remain fully in control,” the spokesperson said. “AI agents can only initiate purchases within clearly defined, user-set parameters, such as spending limits, approved merchants, or required approvals.”
Visa said in a press release that it’s planning for other AI-driven tools and features, like an Agent Score rating system for merchants to evaluate whether or not their website is ready for AI commerce. There’s also an Agentic Directory coming for merchants to know which AI agents to trust for transactions, including Visa-verified agents and merchants. The company also plans to use an AI model to improve fraud detection.
We’ll have to wait and see when this feature rolls out, where it’s available first and what its limitations are.
Watch this: AI Data Center Infrastructure Plans Continue to Draw Controversy
The problem with giving AI your payment information
These security measures don’t necessarily guarantee you won’t fall victim to fraud or theft. And Visa and OpenAI haven’t shared who would be responsible if the AI agent makes a costly mistake. Imagine trusting AI to buy your luggage, but choosing the wrong one or shipping it to the wrong address. The companies have not responded to a request for comment.
Do people even want this? A Bain & Company 2025 survey found that about a quarter (24%) of US consumers surveyed would feel comfortable letting AI make a purchase for them. If most US adults aren’t fans of bots shopping and buying for them, why enable it?
AI is being positioned as a helpful tool to save us time, but that convenience comes at a cost when our personal information is at risk. And shopping is personal. While AI may be used to research, compare and recommend, many Americans don’t want a bot acting autonomously on some of their most personal decisions using their hard-earned money.
It’s not OpenAI’s first foray into shopping — and OpenAI isn’t unique among AI companies.
ChatGPT already has ChatGPT Search, so you can shop, compare and buy items directly in the chat without going to an external website. The retired Visa Instant Checkout allowed the AI agent to be your personal shopper for an item. But the big difference for both is that you would complete the purchase yourself. AI-enabled shopping has been highlighted by other companies, like Google, Apple and Samsung, in their pitches about agentic AI.
In the meantime, use AI in a way that you feel most comfortable — even if that means not using it at all.
“There is no pressure to use these AI tools. If you’re not comfortable doing so, don’t, especially if you don’t understand how your data is being secured and what the AI agents can and can’t do with that data,” Flores said.
Stephan is the sports journalist for the Maple Grove Report.
The requests are sent by the web browser to the web server. The web server will pass the request to the application server, which will generate the SQL to be executed in the database.
