Riskified Formally Discloses Agentic Commerce Fraud Risk in SEC Filing — Structural Threat to Chargeback Guarantee Model

Riskified Adds New Risk Category to Annual Report

Agentic Commerce Poses New Fraud and Liability Risks for Riskified's Chargeback Guarantee and Profitability - TipRanks.com

Riskified Ltd. Class A (RSKD) has disclosed a new risk, in the Demand category. The increasing adoption of Agentic Commerce exposes Riskified Ltd. Class A to potential fraud patterns.

Riskified (NYSE: RSKD), a provider of ecommerce fraud detection platforms, formally disclosed new risks associated with the rise of agentic commerce in its annual report (Form 20-F) filed with the SEC on March 6, 2026, categorizing them under "Demand."

The company stated that the proliferation of AI agents could lead to "shifts in liability frameworks" and "reduced chargeback risk for merchants," potentially decreasing demand for Riskified's products and eroding the value of its chargeback guarantee. Furthermore, consumer use of AI agents could increase transaction disputes and friendly fraud, adversely affecting chargeback-to-sales ratios, margins, and AI model accuracy.

Industry Context and Background

What makes this risk disclosure particularly notable is that Riskified itself had announced an expansion of its AI agent fraud detection capabilities just three days earlier on March 3. The juxtaposition of an offensive product launch with a defensive risk disclosure in SEC filings reveals both the business opportunity and structural threat in this space.

Concerns about agentic commerce extend across the industry. According to a Visa report, dark web posts mentioning "AI agents" have increased by over 450% in the past six months. Malicious bot transactions have grown 40% in the US, revealing that fraud organizations are actively studying and exploiting AI agent mechanisms.

Friendly fraud (where legitimate consumers file chargebacks for illegitimate reasons after purchase) is already a global challenge. It accounts for approximately 75% of all chargebacks industry-wide, with annual damages estimated at $132 billion. As AI agent-mediated purchases increase, claims of "I didn't know I bought this" become easier to make, and this problem is expected to intensify.

Structural Challenges Facing the Chargeback Guarantee Model

The core of Riskified's business model is the "chargeback guarantee." The company makes approve/decline decisions on behalf of merchants and compensates for losses when chargebacks occur on approved transactions. This model requires highly accurate fraud detection to remain viable.

However, agentic commerce fundamentally undermines this premise. As Riskified details on its blog, AI agent-mediated transactions significantly reduce the "interaction data" essential for traditional fraud detection. Specifically, signals such as device information, browsing behavior, mouse movements, and source IP addresses are replaced by the agent's uniform communication patterns.

Xavi Sheikrojan, Director of Risk Intelligence at competing fraud detection firm Signifyd, points out:

What has changed is not the motive for fraud, but how it manifests. When agents handle shopping, the human session disappears. Fraud no longer stands out as noisy behavior — it blends into transactions that look perfect.

Source: Xavi Sheikrojan, Signifyd

Even more concerning is the "liability gap" issue. As Riskified's Jeff Otto explains, transactions through OpenAI's Agentic Commerce Protocol can be completed without consumers ever visiting the merchant's site. Even when card fraud occurs, chargeback liability remains with the merchant. For Riskified's chargeback guarantee model, the increase in such "transactions beyond merchant control" makes guarantee costs increasingly difficult to predict.

Impact on Ecommerce Merchants and Countermeasures

This risk disclosure contains several important implications for ecommerce merchants.

First, visibility into AI agent transactions is urgent. As Riskified recommends, merchants need to build identification and monitoring systems for agentic commerce transactions, continuously checking whether transaction volume spikes are due to fraud attacks. Legacy bot detection tools may not be able to distinguish between legitimate AI bots and malicious ones, necessitating solution reviews.

Second, fraud detection must transition to a "network-based" approach. Rather than judging individual transaction data in isolation, leveraging "intelligence networks" that cross-reference network-wide purchasing patterns and historical non-agent transaction data becomes crucial.

The most fundamental challenge is the revision of contracts and policies regarding chargeback liability allocation. As AI agent-mediated transactions grow, the debate over who bears risk among merchants, platform operators, and LLM providers is inevitable. Payment networks like Visa are also advancing their response, having invested over $13 billion in security technology over the past five years, but industry-wide rule development is still in progress.

Summary

Riskified's SEC risk disclosure demonstrates that agentic commerce is not merely a technology trend but a "structural change" that alters the premises of ecommerce revenue structures and risk management. The move of taking an offensive stance with a March 3 product announcement while candidly disclosing risks in a March 6 SEC filing speaks to the high uncertainty in this domain.

Key points to watch going forward are when payment networks like Visa and Mastercard will establish rules for agentic commerce liability frameworks, and whether fraud detection providers including Riskified can maintain equivalent accuracy without interaction data. For ecommerce merchants, pursuing the growth opportunities of AI agent adoption while simultaneously redesigning risk management frameworks is now essential.