Sumsub's Identity Fraud Report 2025-2026

Key Insights

• 31% Drop in South African Identity Fraud: South Africa’s fraud rate fell 31% YoY to 1.4% in 2025, a standout amid continental volatility. Attributed to mature AML/CFT frameworks, bank-led verification enhancements, and biometric eKYC rollout, it curbed traditional document/selfie scams. However, this masks emerging threats: deepfake incidents rose +269% YoY, signaling AI impersonation as the next frontier in its digital landscape.
• Growth of Deepfakes: Deepfakes exploded across Africa, embodying the Sophistication Shift. Continent-wide selfie fraud surged, with deepfakes fueling mismatches. Highlights: DRC (+367% YoY), Malawi (+325%), Tanzania (+317%, now 0.71% of attempts), Kenya (nearly 10% of fraud). Globally, they rank 11% of first-party fraud, evolving to interactive avatars bypassing liveness—challenging visual defenses and demanding multi-modal (e.g., audio/device) counters.

The newly released Sumsub Identity Fraud Report 2025-2026 delivers a data-driven analysis of evolving identity fraud trends, drawing from over 4 million fraud attempts processed on its platform in 2024-2025. This is paired alongside surveys of 300+ fraud professionals and 1,200+ end-users across global regions.

The report shows a shift in focus from the rapid growth of fraud”— where fraud-as-a-service (FaaS) tools lowered entry barriers, to the “Sophistication Shift,” a maturation where attacks become fewer but more targeted, costly, and damaging.

As CEO Andrew Sever notes, generative AI has “democratized deception” but compelled verification innovations to match pace, emphasizing human insight, data intelligence, and AI precision for scalable trust.

The methodology aggregates anonymized verification data from high-traffic jurisdictions (over 15,000 attempts each), excluding low-sample areas for reliability. It compares 2024-2025 trends, incorporating 2023 baselines for context, and integrates the 2025 Fraud Exposure Survey insights from sectors like banking, crypto, and e-commerce. Graphs highlight internal stats, revealing a global fraud rate stabilising at around 2.5% but with rising sophistication: a 180% YoY increase in advanced techniques like deepfakes and social engineering.

Key Trends: From Volume to Precision

The report identifies two pivotal trends:

First, the Sophistication Shift filters out “low-effort” scams (e.g., crude forgeries), pushing fraudsters toward strategic operations. Public awareness has been effective in reducing amateur attempts, but evolving tech has meant sustained challenges. Fraud now demands greater investment, yielding higher impacts like prolonged account abuse.

Second, AI industrialises fraud, evolving from isolated tools to an entire ecosystem. Document forgeries leverage OpenAI-like content generators for flawless holograms and fonts; This is a scenario where watermarks are easily stripped. Synthetic videos via Google Veo or OpenAI’s Sora 2, create dynamic deepfakes with micro-expressions, bypassing even sophisticated liveness checks. This “fraud production line” enables professionalized rings, with AI agents—autonomous systems blending generative models and reinforcement learning, building orchestrated end-to-end attacks by fabricating identities, and enabling interacting in real-time, while adapting rapidly from failures.

Global Fraud Landscape

Fraud rates have hovered at a steady rate, but AI composition abilities has diversified into first-party (self-perpetrated, e.g., application fraud) and third-party (impersonation, e.g., account takeover) categories. First-party tops include synthetic identity use (21%), chargeback abuse (16%), and deepfakes (11%), per survey data. Third-party leaders: identity theft (28%), account takeover (19%), and card testing (17%). ID cards dominate fraud share (72%), followed by passports (13%) and driver’s licenses (10%).

Regionally, hotspot growth varies: Middle East (Iraq 9.7%), APAC (Pakistan 5.9%), Africa (Tanzania 5.0%). Industries face uneven threats. Dating and online media hit 6.3%, crypto 2.2%.

Key findings spotlight vulnerabilities: Iraq’s ID card fraud at 10.2%; Maldives’ deepfake surge (+2,100% YoY); Zambia’s 37% approved applicants in fraud networks; Nigeria’s 8% synthetic documents; Malaysia’s +197% fraud growth.

The report’s AI/digital fraud breakdown warns of “agentic AI”: self-executing fraud chains, from persona creation to telemetry tampering. By late 2025, state-linked espionage used such agents for phishing without human input, a foreshadow of financial adaptations. Fraud defences have pivoted to behavioural AI. This includes the onboarding of flow scripting detection, transaction geolocation monitoring, and cross-signal fingerprinting to expose rings. Multi-modal liveness (vision, audio, device) counters deepfakes, while structure-aware models validate document logic (e.g., MRZ checksums).

Industry Breakdowns

Financial services lead at 2.7% fraud, driven by application and ATO schemes. Crypto (2.2%) suffers synthetic identities; professional services (1.6%) sees +232% YoY from freelance platforms. Dating/online media (6.3%) battles social engineering; iGaming/e-commerce face mule proliferation.

Regional Focus on Africa

Africa exemplifies the Sophistication Shift amid digital boom with mobile money and fintech fuel inclusion but has also created fraud battlegrounds. Selfie fraud has surged, with mismatched/manipulated face usage up dramatically, tied to deepfake spikes. Overall rates of fraud is a mixed patchwork, rising in West Africa hubs, but seeing drop-off in well regulated markets.

Top fraud countries: Tanzania (5.0%), Cameroon (4.8%), Mali (4.6%). YoY growth leaders: Mali (+131%), Lesotho (+88%), Mozambique (+77%).

Declining regions: Nigeria (-54% to 2.7%), Algeria (-60% to 2.3%). Fraud networks concentrate in Zambia (37% approved applicants linked), Rwanda (34%), Malawi (27%).

Surveys reveal 76.5% awareness of money muling, but 1 in 4 targeted underestimate risks, mistaking recruitment for jobs. Deepfakes integrate into playbooks, often mislabeled as mismatches due to realism.

Fraud Forecast for 2026

AI’s double edge sharpens: fraudsters automate vulnerabilities; defenders leverage behavioral signals. Biometrics + AI agents enable delegated tasks (e.g., shopping), blurring synthetic lines—requiring agent verification. Mule multiplication hits high-volume sectors; countermeasures need transaction monitoring and intelligence sharing. Social engineering exploits digital literacy gaps, boosting APP/ATO in vulnerable areas. Non-Doc Verification surges (+338% adoption), database-backed to sidestep forgeries—popular in Canada/Singapore, setting compliance benchmarks.

Fragmented stacks yield to unified workbenches blending compliance/fraud with AI. Fraud/compliance convergence forms risk units for lifecycle oversight.

Winning Fraud Prevention Strategy

The report advocates layered defences: Behavioural AI for adaptation, Non-Doc for frictionless compliance, and ecosystem integration for agility. Sumsub aids fraud defences via its platform, offering AI-driven verification, mule detection, and regulatory tools.

In sum, 2025’s shift demands proactive evolution, by treating fraud as an arms race where velocity trumps volume. As Pavel Goldman-Kalaydin (Sumsub AI/ML Head) states, AI verifies “not just who you are, but who acts on your behalf.”

Comparison: Identity Fraud Rates in Select African Countries (2025)

This table draws from report charts (10-12), focusing on top/highlighted countries. Declines in regulated markets like South Africa contrast rises in emerging hubs, Highlighting uneven digital maturity in the African fraud resillience landscape.