Deepfake detection upgrade for Sumsub highlights continuous self-improvement

Sumsub has launched an upgrade to its deepfake detection product with instant online self-learning updates to address rapidly evolving fraud tactics.

A blog from the London biometrics firm says its Adaptive Deepfake Detector is “a new model that tackles the prevailing issue of traditional offline solutions being incapable of catching the newest deepfake scams.” It does via non-stop machine learning from fraud signals across multiple layers, allowing it to adapt within hours when new varieties of threat are detected. That means no waiting time for scheduled training cycles and no need for regular human review.

“In 2026, the threat landscape has evolved, demanding risk management teams to respond with the next-generation fraud prevention models,” says Nikita Marshalkin, head of machine learning at Sumsub. “Modern deepfakes can no longer be detected by the human eye, and decision-making should be based on multiple signal analysis in real time.”

The product’s continuous learning analyzes documents, geolocation, IP address, device signals, facial biometrics for liveness data, and verification information from multiple users to more quickly recognize emerging fraudulent activity, including new deepfake types or injection methods.

Marshalkin says Sumsub’s deepfake detector now offers clients “not just a tool, but rather an online learning system that combines advanced document checks, device intelligence, and fraudulent networks analysis to complement deepfake detection capabilities.”

“When the price of failure is too high, a comprehensive approach to the increasing AI-driven fraud challenge is the answer we need.”

Indonesia’s new regulations demand activity-based compliance

In a new whitepaper, Sumsub argues that continuous, activity-based compliance must become the standard for payment providers in Indonesia, as they navigate new regulations.

Bank Indonesia’s BI Regulation 10/2025, says a Sumsub blog, restructures the regulatory framework governing payment service providers across Indonesia. The whitepaper “serves as a reference for industry players seeking to understand Indonesia’s regulatory framework for payments.”

“Technology-enabled compliance is becoming non-negotiable as the country moves towards the resilient, secure, and integrated payments infrastructure outlined by Blueprint Sistem Pembayaran Indonesia 2030 (BSPI 2030),” it says.

BSPI 2030 is a scheme that aims to oversee digital technologies like APIs, foster business partnerships, and regulate ownership in the payment system. It makes regulatory requirements dependent on specific activities provided.

As such, compliance requirements are now tied directly to specific activities such as issuing e-money, operating a payment gateway, or facilitating fund transfers.

“Indonesia’s payment sector is navigating a structural shift that demands more than a compliance checklist,” says Penny Chai, vice president of Sumsub for APAC. “Point-in-time controls simply cannot keep pace with persistent levels of fraud that are increasingly being strengthened with AI. Sustained growth in the digital era is only possible when businesses stop treating verification as a hurdle and start viewing it as the foundational infrastructure for digital trust. What payment providers need to operate with confidence is a multi-layered defense that works as a continuous system, connecting identity verification, transaction monitoring, and AML screening across the entire customer lifecycle.”

Paper highlights key trends for payments in Indonesia

Sumsub’s whitepaper identifies three key trends shaping the future of payments in Indonesia, which recorded the second-highest fraud rate in APAC in 2025.

First is a shift toward continuous compliance models, which enable Indonesian providers to “move away from fragmented, point-in-time checks to an integrated verification lifecycle.”

Second is low-friction design, providing the near-instant onboarding Indonesians expect – without sacrificing security. “By assessing fraud signals such as device behaviour and network patterns at the point of entry and updating risk profiles dynamically, providers can fast-track low-risk users while ensuring compliance with BI Regulation 10/2025 for higher-risk activities.”

Finally, behavioral intelligence is emerging as a shield against mule networks. “According to Sumsub research, 1 in 4 individuals in APAC have been personally targeted for mule activity. Behavioral analysis and network-based detection are essential to identify complex schemes like mule account networks and micro-transaction layering”

Article Topics

AI fraud  |  biometric liveness detection  |  biometrics  |  continuous authentication  |  deepfake detection  |  deepfakes  |  Sumsub