Visa and Akamai Take Aim at the Trust Gap in AI Commerce

As AI agents begin to execute real actions online, payments and security companies are addressing a concrete operational problem: most e-commerce systems cannot reliably distinguish a legitimate AI agent acting on behalf of a consumer from malicious automation. Visa and Akamai Technologies just announced a partnership aimed at addressing that gap, combining payment-level agent authentication with edge-based traffic analysis to help merchants identify and evaluate AI-driven activity before it reaches checkout and payment systems.

The collaboration links Visa’s Trusted Agent Protocol with Akamai’s behavioral intelligence, bot detection, and user recognition capabilities. Visa said the protocol allows AI agents to securely convey who they represent and what they are authorized to do, while Akamai’s edge platform evaluates behavioral signals associated with that activity as it enters merchant environments. The companies framed the effort as a way to give merchants explicit controls over whether agent-initiated requests are allowed to proceed.

Patrick Sullivan, Akamai’s chief technology officer for security strategy, said the central issue is recognition rather than automation itself. “The promise of agentic commerce hinges on recognition: the fundamental ability to trust an agent acting on someone’s behalf,” he said. Visa’s chief product and strategy officer, Jack Forestell, said the protocol is intended to integrate with existing web and payment infrastructure rather than require new checkout architectures.

Agentic Commerce Exposes Gaps in Fraud Controls

Agentic commerce refers to software agents that can search, compare, and complete transactions autonomously for users. While the concept has gained visibility alongside advances in large language models, merchants and security teams are already contending with its practical implications because automated traffic is rising across retail and payments environments.

Akamai’s 2025 fraud and abuse research documented a roughly 300% year-over-year increase in AI-driven bot traffic, with commerce identified as one of the most targeted sectors. The company said this growth is complicating efforts to separate benign automation from abuse, particularly as tools capable of mimicking human browsing behavior become more accessible.

Traditional bot-mitigation systems were designed to block automated behavior outright. That approach becomes less effective when legitimate AI agents are expected to browse sites programmatically, compare prices at speed, and initiate transactions. Without a way to authenticate the agent and link it to a verified consumer, merchants face a choice between blocking automation entirely or accepting higher fraud risk.

Visa’s Trusted Agent Protocol is intended to address this by allowing agents to present cryptographic credentials that identify both the agent and the consumer it represents, along with signals describing the scope of permitted actions. Visa said the protocol relies on existing web standards so merchants can evaluate those credentials without redesigning checkout or payment flows.

Akamai’s role is to assess those interactions at the edge of the network, before traffic reaches merchant applications. Akamai operates a globally distributed edge platform already used by large retailers to inspect traffic patterns, evaluate request behavior, and enforce bot-management policies. In the context of agentic commerce, those same systems are being used to preserve identity and behavioral context as requests move downstream.

How Payments and Infrastructure Providers Are Responding

The Visa-Akamai partnership is part of a broader set of responses from payments networks and infrastructure providers confronting similar operational challenges. Visa introduced the Trusted Agent Protocol earlier in 2025 and has said it is working with ecosystem partners, including Cloudflare, Stripe, Microsoft, Adyen, and Shopify, to support agent authentication across different parts of the commerce stack.

Other payments companies are taking comparable steps. Mastercard announced Agent Pay in 2025, describing it as a framework that allows AI agents to initiate transactions within existing authorization and compliance structures rather than bypass them. The company positioned the product as an extension of its current acceptance network rather than a new payment rail.

Stripe has invested in machine learning across its payment stack for years, and this year, the company announced a Payments Foundation Model that it says has been trained on tens of billions of transactions to capture nuanced signals about payments and improve performance. Early deployment of the model has been credited with materially increasing fraud detection rates for certain attack patterns compared with prior systems, illustrating how AI is already embedded in live payment flows at scale.

For merchants, the Visa-Akamai collaboration adds tooling at two control points already used in production environments. Akamai’s edge platform evaluates behavioral signals before requests reach merchant systems, while Visa’s protocol provides a standardized mechanism for agents to present verifiable credentials tied to a consumer. The companies said this allows merchants to make explicit decisions about whether to allow, restrict, or block agent-initiated activity using identity and risk signals they already rely on.

Rather than introducing a new consumer-facing shopping model, the partnership addresses how automated traffic is authenticated and governed within existing commerce infrastructure. In that sense, the work around agentic commerce is unfolding less as a shift in how people shop and more as an extension of long-running efforts by payments and security providers to manage fraud, identity, and authorization as automation becomes more capable.