Deepfakes rise 31% as identity fraud grows more complex

Deepfakes rise 31% as identity fraud grows more complex

More than half of businesses have already suffered fraud involving AI agents, yet only a small fraction can cut off an agent’s access once something goes wrong, according to new research from Sumsub. The findings come from the company’s sixth annual Identity Fraud Report 2026–2027.

The global identity fraud rate dropped 28% year on year to 1.6%. Lower volumes are masking a tougher threat, however. Attacks are becoming more organised, harder to spot and increasingly powered by AI, with deepfakes rising 31% worldwide. Sumsub argues that fewer attempts do not mean lower exposure.

As criminals adopt sophisticated tools, work through linked networks and exploit AI-driven processes, firms must do far more than flag suspicious users or risky payments.

Agentic fraud stands out as one of the most pressing new risks. Sumsub’s Fraud Exposure Survey 2026 found that 53% of businesses have been affected by fraud involving AI agents. Only 37% monitor who deploys these agents, and just 12% are able to suspend or revoke their access. Companies are letting software act for staff without any reliable way to confirm its identity, check its authority or remove its permissions.

Coordinated fraud networks present a second major blind spot. Some 74% of companies fail to detect network-level warning signs, and only 21% use fraud network analysis. Among firms that experienced fraud, a third said at least 25% of confirmed cases were tied to a broader network.

Links can surface through shared devices, contact details, payment methods, locations, IP patterns or document templates, and onboarding checks rarely catch them.

Sumsub’s own data shows that roughly 2% of verified users are later connected to a fraudulent cluster, some spanning several countries and involving thousands of accounts. Money mule schemes, where legitimate accounts are sold or passed to criminals to move illicit funds, illustrate the problem.

Regional trends diverge sharply. APAC logged nearly one million confirmed fraud attempts, the most of any region, while Japan’s fraud rate climbed 219% to 5.1% of verifications. The US recorded the largest single fraud network at 3,173 users, even as North American fraud rates fell 41%.

Latin America showed the highest consumer exposure at 85%. In Africa, Algeria saw the world’s steepest deepfake increase at 1,205%. The Middle East saw deepfakes surge in 11 of 13 markets studied, and Europe recorded a 41% rise in deepfakes despite an 8% fall in fraud rates.

Sector patterns also varied. Financial services, crypto platforms and online media saw fewer incidents but greater complexity, while e-commerce and IT services worsened on both counts. Dating apps posted the highest fraud rate alongside the lowest consumer trust.

Looking to 2027, Sumsub expects growth in privacy-preserving age verification, mechanisms to verify AI agents and their human owners, network-focused detection, verification methods that do not rely on identity documents, and layered, risk-driven controls that avoid adding friction for genuine users.

Sumsub co-founder and CEO Andrew Sever said, “We see a fundamental shift in digital fraud: while fraud rates are declining in some regions, threats are becoming more sophisticated, interconnected and harder to detect across all regions. AI is accelerating this evolution, exposing the limitations of traditional account- and transaction-focused fraud prevention.

“Businesses aiming to respond effectively to this increasingly complex threat landscape must prioritize building a unified trust infrastructure that combines verification and compliance needs with advanced fraud detection and risk intelligence”

Sumsub CTO and co-founder Vyacheslav Zholudev added, “Automation is neither inherently good nor bad; in fact, agentic software is becoming an essential link between individuals and their digital information. However, this growing convenience must be matched by effective oversight”

He added, “With AI Agent Verification, Sumsub is the first to bind AI agents to verified human identities at scale. The recent solution upgrade also enables teams to control what data is available to the agents at any given moment, making agentic deployment transparent, traceable and customizable according to the business needs”.

Read the daily RegTech news

Copyright © 2026 RegTech Analyst

Enjoyed the story? 

Subscribe to our weekly RegTech newsletter and get the latest industry news & research

Copyright © 2026 RegTech Analyst

Investors

The following investor(s) were tagged in this article.