When AI Does the Shopping, Who Takes the Risk?
Click here to view/listen to our blogcast.
Imagine telling an AI assistant, “Find the best deal on a laptop for my office, compare warranties, and order five.” No twenty browser tabs, just a recommendation and a purchase. Sounds great. But what happens if it orders the wrong model, exposes your credit card, or buys from a fraudulent seller? And what happens to the websites it visits?
The debate is already underway. On November 11, 2026, United Airlines announced restrictions on unauthorized third-party AI agents accessing its booking systems. Just weeks earlier, Southwest Airlines introduced a ChatGPT integration that helps travelers find and compare flights before completing purchases on Southwest’s website.
Two airlines, two approaches.
Why Businesses Are Concerned
Unlike a search engine, an AI agent can take action: signing in, changing reservations, or purchasing items with your permission. That creates several risks:
- Privacy and security: An agent with access to accounts, payment details, or customer records could expose sensitive information if compromised or manipulated.
- Unintended actions: The wrong dates, an incorrect quantity, or an accidental use of reward points can become an expensive mistake.
- Regulatory obligations: Some transactions require explicit human acknowledgment. United, for example, cites airline check-in declarations concerning prohibited items in luggage.
- Fraud and system abuse: Businesses must distinguish authorized customer assistants from malicious bots, automated account attacks, and traffic that strains their systems.
Another threat is prompt injection: malicious instructions hidden on a website that trick an AI agent into acting against its user’s wishes. OpenAI’s security research explains why this remains a significant challenge even for sophisticated assistants.
The Price Comparison Problem
Here’s the part that isn’t strictly about security: AI agents could change how companies make money.
Today, retailers pay to attract visitors, display sponsored products, recommend accessories, and encourage customers to spend more. An AI assistant asked to find the lowest total price may bypass much of that experience. It can compare competing stores, shipping charges, return policies, and specifications without ever admiring a sponsored display or impulse-buying a matching accessory. Shoppers benefit. Retailers that depend on advertising, loyalty, and add-on sales may not.
A November 4 Reuters analysis explored this tension at Amazon and Walmart, which are taking different approaches to third-party AI shopping. Blocking agents can protect more than customers; it may also protect a company’s ability to shape what customers see.
Not every restriction is an attempt to avoid competition. And AI comparisons are not necessarily impartial. An assistant may miss discounts, overlook fees, rely on outdated information, or favor products that are easiest for it to access.
Could Blocking AI Backfire?
For a small retailer, hotel, or service business, shutting out AI assistants could mean shutting out potential customers. If buyers increasingly ask agents to shortlist products and suppliers, businesses absent from those results may never get considered. The choice is not simply “allow every bot” or “block all AI.”
Shopify now lets merchants manage participating AI shopping channels, including where product information is shared and whether supported purchases can happen within an AI experience. An AI crawler that reads public product information is also different from an AI agent attempting to access customer accounts or complete transactions. They should not automatically receive the same permissions.
Who Is Responsible When Something Goes Wrong?
Suppose an assistant books a nonrefundable room for the wrong night. Is that the customer’s mistake, the AI provider’s, or the hotel’s? The answer depends on the agreement, the authorization given, and applicable law. It isn’t settled by saying, “The AI did it.”
Businesses need clear rules for automated visitors. Customers need to understand what they authorize.
What You Can Do
If you use AI assistants:
- Start with research and comparisons. Review actual prices, fees, terms, and sellers yourself.
- Use approved integrations rather than sharing account passwords whenever possible.
- Require your own confirmation before purchases, transfers, or account changes.
If your organization operates a website or online store:
- Review whether your current bot protections distinguish useful AI discovery from unwanted automation.
- Keep logins, payment systems, and sensitive transactions behind appropriate safeguards and human approval.
- Consider controlled, authorized AI integrations, and test what agents actually see about your products or services.
How CDML Can Help
AI access is becoming another business-security decision, much like deciding who can access your network or customer records. CDML Computer Services can help organizations review access controls, protect accounts, monitor suspicious activity, and develop sensible policies for using AI tools without giving them unnecessary authority.
Final Thoughts
AI agents could make buying faster and more competitive. They could also introduce security risks and pressure on online businesses. The question isn’t whether AI should help us shop. It’s how much authority we’re willing to give it, and who accepts responsibility when it acts on our behalf.
If your organization is evaluating AI tools or automated access to its systems, contact CDML Computer Services to discuss a practical approach.
Stay safe. Stay informed. Stay compliant.

📞 Contact us here: https://cdml.com/contact/
📚 Read more on our blog: https://cdml.com/blog-2
📺 Listen to our blogcasts: https://www.youtube.com/@CDMLComputerServices


