Recently, six global banks warned that as AI shopping agents take on more responsibility for purchases, the industry must address privacy, fraud, transparency, and customer protection – especially ahead of the holiday season. Thales, which protects 70% of the world’s 50 largest retailers and 92 of the 100 largest global banks, recently released data on the AI threat facing the retail/e-commerce industry.
Todd Moore, Global VP of Data Security at Thales, said 72% of retail/e-commerce leaders say AI is changing too quickly for security teams to keep pace.
Sensitive data is at greater risk: 61% report AI-driven sensitive-data disclosure; only 37% have a complete picture of where their data sits, according to Thales.
Retailers need clear accountability: Before agent-led shopping becomes routine, brands must define who is responsible when an automated purchase, recommendation or data exchange goes wrong, added Thales.


Powerful new tools
AI shopping agents are a powerful new tool for retailers and consumers, said Moore.
“Shoppers can easily search, compare, and purchase products, while retailers can offer suggestions and deliver a more personalized experience. However, the main security risks of these agents is making sure they have clear authorization to take actions on behalf of consumers and are operating within set limits,” he said.
“Ahead of the holiday shopping season, retailers should focus on establishing strong identity verification systems, writing granular permissions that AI tools can’t override, and introducing confirmation steps for high-risk actions that require a human’s sign-off. In situations where an AI agent needs to review a customer’s purchase history or payment credentials, retailers need to make sure the agent can only access what it needs and that a customer can see and revoke permissions at any time. Convenience cannot come at the cost of control.”
Keeping pace with security
Given the ease of access for many of today’s AI tools, retail security teams do not have a complete inventory of what AI applications, agents, data sources, and third-party connections operate within their environment, said Moore.
“This makes it hard to know who or what has access to sensitive data like payment workflows and personal customer information,” he said.
“The goal of any attacker is to achieve their goal with the least amount of effort. For consumer attacks, this means looking for exposed customer profiles, purchase histories, loyalty records, addresses, authentication data, payment tokens, and customer-service conversations. When targeting retailers, the most appealing data is internal credentials, pricing strategy, inventory data, supplier terms, and sales forecasts.
“As for securing these data sources, retailers should start with data discovery and classification. Our DTR (Direct to Retail) shows that only 37% of retail and e-commerce organizations know where all their data is stored, so it’s impossible for them to protect what they can’t locate. From there, retail security teams can apply least-privilege access and prevent sensitive inputs from being used to train or enrich AI systems without explicit controls.”

Establishing clear accountability
Accountability should be visible throughout all stages of the transaction, not just after an incident occurs, added Moore.
“Every agent a retailer uses should provide a clear record of who authorized it, what information it accessed, what limits applied, what decision it made, and which party was responsible at each stage. Once something happens, retailers need the ability to pull auditable logs and implement a clear incident-response and customer-remediation process,” he said.
“Trust has become a conversion metric that leads to whether or not a consumer purchases a product using an agent or not. Consumers may use an AI agent to research a product, but they will not delegate purchasing decisions to it if they don’t understand what the agent is doing, what data it is using, or how they can intervene. Retailers should make these controls visible to gain consumer trust in the AI tools they deploy.”
More from Retail Insider:












