Telling Friend From Fraud In The Agentic AI Economy

InnovationTelling Friend From Fraud In The Agentic AI EconomyByAndrew Sever,Forbes Councils Member.for Forbes Technology CouncilCOUNCIL POSTExpertise from Forbes Councils members, operated under license. Opinions expressed are those of the author. | Membership (fee-based)May 28, 2026, 08:15am EDTAndrew Sever is the CEO of Sumsub, a leading full-cycle verification platform that enables scalable compliance. getty​With 85% of enterprises now exploring agentic AI, a new economy centered around the use of AI agents has dawned. However, the current security models built around assuming a human is always behind an action are fast becoming outdated and opening up opportunities for abuse.In financial institutions, for instance, systems have long been built around the assumption that a human is initiating a transaction. Any signs pointing to potential automation would thus be a red flag. Throughout this process, the intent and execution of the action are assumed to be bound closely together in one and the same entity.This assumption, however, is being challenged as intent and execution are diverging. Today, AI agents can perform virtually any task that can be done on a browser, often without continuous human input. Users can set the intent of an action and assign an AI agent to automate it on their behalf, creating new vulnerabilities and marking a new era for trust and accountability.Automation Versus Authorization​Automation itself isn’t the problem. The challenge is that it’s becoming more accessible for both legitimate and malicious use cases. While AI agents can help in tasks like making one-off mass payments or purchasing multiple tickets, they also make it significantly easier for fraudsters to industrialize attacks in a way that once required far more technical expertise. There’s an urgency to make sure that the entity behind an automated action is legitimate and authorized to perform it.Good Agents, Bad Agents And Staying Accountable​This leads to an uncomfortable am...