US-based Riskified has announced an expansion of its AI Agent Intelligence platform, aimed at providing fraud and risk protection for native AI shopping assistants deployed by merchants on their digital storefronts.

This move comes in response to the increasing investment by retailers in conversational AI agents as customer-facing tools. According to recent research from McKinsey & Company, 82% of retail organizations have already launched generative AI pilots focused on reinventing customer service.

Enhanced Capabilities for Merchant AI Agents

The expansion introduces two key capabilities designed specifically for merchants running their own AI shopping assistants.

First, AI Agent Identity Signals lets a merchant’s AI shopping agent directly query Riskified’s Identity Graph to retrieve real-time risk indicators and resolve customer identities. This functionality is accessible through Riskified’s Model Context Protocol (MCP) integration on AWS Marketplace, via Google’s Agent-to-Agent (A2A) protocol, or through standard RESTful APIs. It enables instant decision-making, such as approving refunds or exchanges, based on consumer risk and eligibility during live interactions.

Secondly, the AI Agent Policy Builder is an improved version of Riskified’s Decision Studio product. It allows merchants to define and apply business rules to manage the risk profile of order volumes originating from their native AI agents, thereby providing a mechanism to limit exposure to issues like refund claim abuse, reseller arbitrage, and promotion misuse.

Assaf Feldman, CTO and co-founder of Riskified, stated that the company’s role is to serve as the risk intelligence layer that secures and enhances AI agent interactions. This involves augmenting a merchant’s proprietary customer history with insights from Riskified’s multi-merchant network to ensure that transactions and claims processed through AI agents involve verified identities.

Mitigating Fraud Risks

Riskified highlights the growing threat of fraud rings exploiting early agentic protocols and chatbots. The company’s approach focuses on identity resolution and detecting anomalies in purchasing patterns across its global network of e-commerce brands, a capability that individual merchants cannot replicate alone.

The platform is designed to ensure that AI-mediated shopping remains a revenue channel rather than a source of financial exposure for merchants.