European deepfake fraud rises despite broader decline

European deepfake fraud rises despite broader decline

European deepfake fraud increased sharply while overall detection rates declined. Sumsub’s research identifies increasingly complex attacks and limited organisational control over AI agents.


Detected deepfake fraud increased by 41% across Europe in 2026, even as the region’s overall identity fraud rate declined, according to research examining the changing methods used to compromise digital identity systems.

The findings come from Sumsub‘s Identity Fraud Report 2026–2027, which combines information gathered through the company’s identity verification platform with surveys of businesses and consumers.

Across Europe, the overall detected fraud rate declined by 8%, while the technical complexity of fraud cases increased by 13%.

Globally, Sumsub recorded a 28% decline in the average identity fraud rate, from 2.2% in 2025 to 1.6% in 2026. Deepfake-related fraud increased by 31% over the same period.

The contrasting trends indicate a change in the composition of fraudulent activity detected by the provider. Fewer verification attempts are being classified as fraudulent overall, but more of the activity involves technically sophisticated methods.

Deepfakes use artificial intelligence to create or manipulate images, video, and audio, potentially allowing fraudsters to impersonate individuals during remote verification or other digital interactions.

For organisations providing financial services, digital marketplaces, and other online accounts, such techniques can complicate processes that depend on verifying a customer’s identity before granting access.

Sumsub also identified increasingly coordinated fraud operations involving groups of accounts connected through devices, contact information, locations, payment instruments, or other shared characteristics.

These networks can be difficult to identify when security systems examine individual accounts or transactions without assessing relationships between them.

Businesses report weaknesses in AI-agent controls —

Separate survey findings suggest that organisations are encountering another security challenge as they introduce AI agents capable of acting within business systems.

Among business respondents, 53% reported experiencing fraud involving AI agents. However, only 37% said their organisations tracked who deployed agents, while 12% reported having the ability to suspend or revoke an agent’s access.

Those percentages represent the experiences and capabilities reported by survey participants. They are not verified fraud rates derived from Sumsub’s identity verification platform.

The distinction is important because reports of agent-related fraud can encompass different forms of activity, while the platform’s verification data measure detected identity fraud within its own services.

AI agents can interact with applications, retrieve information, and initiate tasks on behalf of individuals or organisations.

Where an agent is permitted to access sensitive records or operational systems, its identity and permissions become part of the organisation’s wider security arrangements.

Businesses need to distinguish between a human user’s permission to perform an activity and the authority delegated to automated software.

An agent may be deployed legitimately but subsequently receive manipulated instructions, act outside its intended role, or retain access after that authority should have been withdrawn.

The survey also identified limitations in organisations’ ability to detect coordinated fraud. Sumsub reported that 74% of surveyed businesses missed network-level fraud signals, while only 21% used fraud network analysis.

Among organisations that had experienced fraud, one-third reported that at least a quarter of confirmed cases were connected to a wider network.

The company found different patterns between industries. Financial services and online media experienced declining fraud rates alongside greater complexity, while e-commerce and IT services recorded increases in both measures.

These findings reflect activity observed within Sumsub’s customer base and the separate survey population, rather than a comprehensive assessment of every business or identity transaction worldwide.

The report draws on 5.8 million fraud samples collected during 2025 and 2026, alongside responses from 1,400 risk professionals and 1,500 consumers. Geographic analysis includes jurisdictions with at least 15,000 verification attempts.

Across those datasets, the findings indicate that organisations face growing challenges in identifying linked fraud activity and maintaining oversight of automated systems, even where the overall proportion of detected identity fraud is declining.

—