Opinion
A new dawn in digital fraud threats: The rise of agentic AI
The emergence of Agentic AI shopping agents is introducing new fraud risks and disrupting the delicate balance between innovation and secure commerce. Eyal Elazar, VP of Market Intelligence & Product Marketing at Riskified, outlines the current landscape and explores potential strategies for addressing these evolving threats.
By 2025, approximately 2.8 billion people around the world will be shopping online. According to Statista, around 21 percent of retail purchases currently take place on the internet, a figure expected to grow to 22.6 percent by 2027. Alongside the opportunities in digital commerce, fundamental changes are taking place in the way consumers shop online, driven by the rise of AI-based shopping agents (Agentic AI). Consumers are increasingly using language-based AI tools (LLMs) such as ChatGPT, Claude, Gemini, and Perplexity to research products, compare prices, and even complete transactions. But while these tools open up endless opportunities, they also increase the risk of a future where fraud becomes more challenging, sophisticated, and frequent.
The Future of Shopping
The entry of AI agents is reshaping the rules of the game for ecommerce merchants. On one end of the spectrum, if this evolution continues naturally, agents may turn the shopping experience into a purely mathematical process. Users might rarely visit websites themselves, leaving the agents to handle all purchases. On the other end, merchants aiming to maintain brand loyalty are building AI-based shopping experiences that preserve interaction and control over the customer journey, enabling them to face the changes head-on. In the middle lies a hybrid approach, where merchants collaborate with external agents that assist with purchases without being blocked by security systems, while still providing the merchant with insight into inventory and key data needed for smart decision-making.
Fraud risk management is becoming more complex. Technology is making it harder to detect and prevent fraud, and cybercriminals are leveraging tools like WarmGPT to create synthetic identities and carry out sophisticated attacks. One example is automated purchases using stolen credit cards that bypass standard fraud detection systems. Data shows that traffic originating from LLMs is twice as risky compared to other sources, highlighting the urgent need for intelligent fraud detection systems and proactive preparation while still maintaining a safe and smooth shopping experience.
Digital Threats Up Close
A recent high-profile case illustrates the danger. Researchers from Zenity demonstrated at the Black Hat conference in the U.S. an advanced exploit of ChatGPT that allowed an attacker to take control of user accounts, access conversations, and retrieve personal files from Google Drive, all without any user interaction. This type of attack, classified as “zero-click,” could potentially endanger millions of users.
In 2024, digital fraud crossed a new threshold in both scope and financial damage. According to the U.S. Federal Trade Commission (FTC), reported losses from various types of fraud reached 12.5 billion dollars. One in three incidents resulted in a financial loss, marking a significant increase compared to the previous year.
While AI shopping agents help merchants attract new customers, they also pose new threats. These agents can execute massive automated purchases, maliciously exploit promotions and coupons, and undermine pricing strategies and customer trust. So how can merchants embrace this technology without exposing themselves to significant risk? This is where intelligent real-time fraud detection systems, agent behavior monitoring, and protocol verification come into play. These tools enable merchants to stay a step ahead of fraud while preserving a safe shopping experience.
Tech-Driven Fraud Prevention
Today’s advanced technologies, based on machine learning and real-time behavioral analytics, are becoming essential tools for identifying emerging fraud patterns unique to AI agents. These systems enable merchants to monitor orders based on their source, whether ChatGPT, Gemini, Claude, or Perplexity, and classify risks with greater precision. This leads to smarter, more controlled decision-making. New verification protocols for AI agents focus on risk management, trust, and data-driven decision-making. They allow for the reliable identification of legitimate AI agents and the prevention of fraud and impersonation.
To fully realize the benefits of AI shopping agents while maintaining strong protection, ecommerce merchants must integrate innovation with advanced security. This means adopting intelligent fraud prevention systems that rely on real-time machine learning and secure agent verification protocols. Collaboration between risk management and business teams enables more precise risk handling, continuity of revenue, and the preservation of customer trust. With this approach, merchants can adapt quickly to the age of autonomous agents and achieve a competitive edge in an ever-changing market.
In this new reality, the same technologies that fraudsters may exploit can also serve as powerful defense tools. With strategic investment in innovation and intelligent systems, AI agents can make digital commerce safer, more trustworthy, and better suited to the era of artificial intelligence, much like how online search was once clunky until Google transformed it. That said, consumers must also remain vigilant. Sharing a credit card or personal information does not eliminate risk, and caution is still needed to avoid phishing and other malicious tactics.
Eyal Elazar is VP, Market Intelligence & Product Marketing, at Riskified.
