Riskified Launches Agent Identity Risk Intelligence to Combat AI-Driven 'Delegated Fraud'

Riskified has launched Agent Identity Risk Intelligence to help merchants distinguish between legitimate consumer AI agents and automated fraud schemes. As personal assistants move from product discovery to autonomous purchasing and returns, fintech professionals must now solve for the identity behind the automation, ensuring that authorized machine-led transactions do not mask sophisticated account takeovers or promotion abuse.

What was announced

The new suite of capabilities is designed to identify the specific AI personal assistant—such as Meta’s Muse, Instinct, or ChatGPT dots—initiating an order or customer service request. By resolving these automated actions to a verified consumer identity, Riskified returns a risk signal in real time. This allows merchants to determine if an interaction, though technically authorized, aligns with the behavior of a trusted customer or a malicious actor. The system addresses a growing shift in commerce where AI agents proactively check for price drops, manage returns, and compare prices across sites. While new standards like the Personal Agent Protocol provide a secure handshake for authorization, Riskified’s tool focuses on the underlying trust of the person granting that authority.

This is particularly relevant for combatting "refund-as-a-service" rings, which use real accounts to run automated fraud at scale. In April 2026, a U.S. federal court sentenced eleven co-conspirators of the Artemis Refund Group, a ring that ran refund fraud as a subscription business against retailers like Amazon, Walmart, and Target. Initially, the capability will be available to merchants on Shopify, where orders from Meta’s Muse are already tagged. Riskified is also integrating this intelligence into Zendesk, enabling both human and AI support agents to see the risk profile of the person behind an automated refund request. Within Riskified Decision Studio, merchants can set specific policies to auto-approve low-risk agent orders while routing high-risk agent-originated refund claims for review.

"Agent protocols solve a real problem: merchants need a secure way to let a customer’s assistant act for them. What they don’t solve is whether that customer should be trusted. A fraud ring can authorize an agent as easily as a loyal shopper can. Riskified sits on top of those protocols and answers the harder question, using what our identity engine already knows about the person behind the agent, so merchants can welcome a good customer’s assistant, decline a bad actor’s, and notice the moment a real customer’s agent starts doing something that the customer never would."

Assaf Feldman, Chief Strategy Officer, Technology and Co-Founder of Riskified.

The companies involved

Riskified (NYSE: RSKD) is a digital fraud and risk intelligence specialist that provides an AI-powered platform to help e-commerce merchants reduce friction and prevent fraud. The company has established a significant footprint in the market by analyzing billions of transactions to build a comprehensive identity engine. This latest move leverages its existing network of order, claim, and chargeback data to address the emerging "agentic" layer of the internet. Shopify, a leading global commerce platform, serves as a primary integration partner for this launch. The platform has increasingly become a testing ground for autonomous shopping features and was an early adopter of the Personal Agent Protocol. Zendesk, another key partner in this rollout, provides customer service software and specialized AI agents. By integrating Riskified’s risk signals into the Zendesk Agent Workspace, the companies are bridging the gap between automated commerce and customer support. These partnerships highlight the interconnected nature of the modern retail stack, where fraud prevention must move beyond the checkout page and into the entire customer lifecycle.

What FF News has reported before

FF News has closely tracked the rise of autonomous commerce and the infrastructure supporting it. We recently covered how Meta and Sierra Launch Personal Agent Protocol with Stripe, Shopify, and Walmart, a move that established the technical standards Riskified is now building upon. The trend toward machine-led shopping has accelerated throughout 2026, with AEON Launches Agentic Checkout and AI Card for Autonomous Shopping on Shopify and Amazon and Juspay Launches Breeze Universal to Enable AI-Powered Agentic Commerce across Platforms. Additionally, we reported on the automation of back-office functions in Datarails Launches AI-Native Ticketing System to Automate CFO Office Workflows, illustrating that the shift toward AI agents is occurring simultaneously in both consumer behavior and corporate operations.

What this means

The industry is entering a phase where "identity" is no longer a static set of credentials but a dynamic relationship between a human and their digital proxies. While the Personal Agent Protocol solved the "how" of agent communication, the burden of "who" falls on fraud prevention layers. This announcement puts pressure on traditional rules-based fraud systems that cannot distinguish between a tireless bargain-hunting bot and a malicious script. The real challenge for the sector will be managing the "gray area" of consumer behavior—where legitimate customers use agents to exploit return policies to the letter—forcing a re-evaluation of what constitutes "fair" participation in digital commerce.