GS interviews Bluefin's Ruston Miles on agentic AI and data protection

As artificial intelligence moves from assisting transactions to initiating them, the payments industry faces a pivotal question: Can security keep pace? In this Q&A, Ruston Miles, founder and chief strategy and development officer at Bluefin, explores what changes when AI agents shop and pay on behalf of consumers. He discusses why traditional point-in-time security models may fall short in an agent-driven environment, what protecting data in motion truly requires, and how payment architectures must evolve to maintain trust as agentic commerce scales.

Green Sheet: What changes when an AI agent, not a human, initiates a transaction, and why does that raise the stakes for data security?

Ruston Miles: When an AI agent initiates a transaction, it introduces sensitive payment and personally identifiable information (PII) into more systems than traditional commerce does. As agents automate transactions and manage identity, the volume of data moving between touchpoints increases significantly.

This raises the stakes for data security because each additional system and handoff expands the risk surface and blurs accountability for protecting that data. Before the payments industry can trust agentic commerce, we need to demonstrate that it can protect data as it moves across touchpoints, not just at the moment a payment is approved. 

GS: You've said trust and data protection are the real limits to agentic commerce. Where do current payment systems fall short?

RM: Most payment systems today are still designed to secure transactions at specific points in time, rather than to manage sensitive data that moves continuously across platforms. Systems protect data during authorization and settlement, but the issue arises when data leaves those boundaries and flows among new merchants, agents, processors and third parties.

Agentic commerce will only expand those connections. Gaps in protection and accountability will become more obvious as transactions are handled by autonomous systems operating across an increasingly interconnected payment environment. This will erode trust when payment systems can't guarantee that data is protected end-to-end. 

GS: What does protecting data in motion actually mean in an agent-driven commerce environment?

RM: In agent-driven environments, sensitive data should never move in a form that can be reused or exploited. Data in this environment doesn't stay in one place or pass through a single transaction anymore. It now moves continuously across touchpoints, and this shift is changing what security needs to do.

Protecting data in motion means designing a security infrastructure that encrypts sensitive information using methods such as tokenization. If data is intercepted at any point in the process, it will have no value on its own. Protecting data in motion is about making security an inherent part of how payments move, not something applied after the fact.

GS: How should security architectures evolve when AI agents operate across platforms, merchants and payment rails?

RM: Security architectures should shift from optimizing for speed to prioritizing trust and transparency. Since AI agents are moving payments and PII across partners and platforms, security cannot be a background thought anymore. Security needs to become the connective layer that enables trusted interactions across networks.

This means protecting payment data and PII at every step of the transaction. Security should power fraud intelligence, enforce consistent controls across channels, and ensure data remains protected as it moves through interconnected systems and payment flows. The goal should not be faster payments with agentic AI. It should enable safer, more secure payments that scale across multiple ecosystems. 

GS: What risks do you see if agentic commerce scales faster than security standards and controls?

RM: If agentic commerce scales before payment security standards mature and evolve, it risks creating agentic systems that move money and PII without clear controls or accountability. With AI agents operating across multiple platforms and payment rails, small security gaps can quickly compound, especially when data is shared across workflows.

This ultimately creates situations where it is unclear who is responsible for protecting data at each step or who is accountable when something goes wrong.

A bigger risk is that these agentic systems are scaling faster than the industry's ability to monitor, audit and contain them. Once agentic transactions become interconnected, a single vulnerability can affect multiple parties simultaneously. 

GS: From your perspective, what must be solved first before consumers can safely trust AI to shop and pay on their behalf?

RM: Agentic AI's potential in commerce is promising right now, but it's too early for secure transactions. Before consumers can trust AI to shop and pay on their behalf, the industry should work to remove sensitive payment data from the transaction flow.

Payments are moving to a model in which merchants never handle card data, using tokenization and embedded platforms to replace static credentials with dynamic data that is useless to attackers. Consumer trust will come when security is built around removing data and shifting responsibility to secure systems, not asking consumers and merchants to manage risk on their own. 

Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.