What's Happening?
More than half (53%) of businesses have experienced fraud involving AI agents, yet only 12% possess the capability to suspend or revoke an agent's access, according to Sumsub's sixth annual Identity Fraud Report 2026–2027. The report indicates that while
the global identity fraud rate decreased by 28% year-over-year to 1.6%, the complexity of attacks has increased, driven significantly by AI, with deepfakes rising by 31% globally. This trend highlights a growing disconnect where fewer fraudulent attempts do not equate to reduced exposure, as fraudsters leverage advanced technologies and interconnected networks. Andrew Sever, co-founder and CEO at Sumsub, noted a fundamental shift in digital fraud, emphasizing that threats are becoming more sophisticated and harder to detect, exposing the limitations of traditional fraud prevention methods. A significant challenge identified is the 'AI authorization gap,' where companies struggle to control non-human actors, with only 37% tracking who deploys AI agents.
Why It's Important?
The findings underscore a critical vulnerability in modern business operations as AI agents become more prevalent. The inability of most businesses to control or revoke access for AI agents creates a significant security risk, potentially leading to widespread and difficult-to-trace fraudulent activities. This 'authorization gap' means that even if fraud rates appear to decline, the impact of each successful attack could be far more severe due to the sophistication and interconnectedness of AI-driven fraud. The rise of deepfakes and organized fraud networks further complicates the landscape, demanding a shift from individual account-focused prevention to a more holistic, network-based approach. Businesses that fail to adapt their fraud detection and risk intelligence strategies to account for AI agents and complex networks face substantial financial losses, reputational damage, and potential regulatory penalties.
What's Next?
The report predicts several key trends for identity and fraud prevention in 2027. These include an accelerated demand for privacy-preserving verification methods beyond basic identity, mechanisms to verify AI agents' identities and their human owners, and a shift in fraud detection towards identifying wider signals and networks rather than isolated incidents. The adoption of non-document verification methods and context-based, multi-layered risk intelligence approaches are also anticipated. Businesses will need to prioritize building unified trust infrastructures that integrate verification and compliance with advanced fraud detection. Sumsub's co-founder and CTO, Vyacheslav Zholudev, highlights the need for solutions like AI Agent Verification to bind AI agents to verified human identities and enable control over data access, making agentic deployment transparent and traceable.
Beyond the Headlines
The proliferation of AI agents and the associated rise in sophisticated fraud present profound ethical and legal challenges. The 'AI authorization gap' raises questions about accountability when autonomous systems commit fraudulent acts. Who is responsible when an AI agent, acting on behalf of a company, causes harm? This issue touches upon the evolving legal frameworks for AI, the need for clear lines of responsibility, and the development of robust governance models for AI deployment. Furthermore, the increasing use of deepfakes and AI-generated content blurs the lines between reality and deception, impacting trust in digital interactions and potentially undermining the integrity of online information. Addressing these challenges requires not only technological solutions but also a re-evaluation of ethical guidelines, legal precedents, and societal norms in an increasingly AI-driven world.













