Estonia-based optical character recognition solutions developer OCR Studio has upgraded its anti-fraud system to detect a broader range of AI-generated and morphed identity documents used during KYC and client onboarding processes. The enhanced system is designed to identify forgeries produced using generative AI models, including ChatGPT, NanoBanana, Grok, and Midjourney.
The upgrade addresses a documented escalation in document fraud enabled by the accessibility and capability of modern generative AI tools. Fraudsters can now replace document photos, generate synthetic identity documents, or reconstruct entire document images at negligible cost and with a level of visual realism that is difficult to detect through manual review.
Scale and financial impact of deepfake fraud
The threat landscape has grown significantly in scale and cost. Interpol has warned that AI fraud schemes are 4.5 times more profitable than traditional methods. According to Surfshark, deepfake-related fraud caused global losses of USD 1.65 billion in 2025 alone, with the US accounting for the largest share of reported losses.
Technical approach
OCR Studio's system addresses these threats through a content-independent approach that examines the underlying image structure of a document rather than searching for logical inconsistencies in its contents. The system detects low-level artefacts left by generation or editing algorithms, allowing it to identify AI-generated forgeries even when they appear visually authentic and to maintain accuracy as generative models continue to evolve.
The company's chief technology officer, Konstantin Bulatov, said generative AI had pushed document fraud to a level where the only detectable traces were invisible to the human eye and undetectable by conventional anti-fraud systems. Businesses that perform customer identification now need specialised deepfake detection technologies not only to protect themselves from fraud, but also to meet increasingly strict regulatory requirements for secure onboarding and identity verification.
Privacy and deployment model
OCR Studio's solutions are deployed on-premise and do not store or transmit identity document data to external services. The company says this approach supports compliance with local and international data protection standards, a consideration that has become increasingly relevant as regulators across multiple jurisdictions tighten requirements around the handling of biometric and identity data during onboarding.
Industry context
The upgrade reflects a broader challenge facing the identity verification sector as generative AI tools become more capable and more widely accessible. KYC processes that rely on visual inspection or pattern-matching approaches designed for traditionally forged documents are increasingly inadequate against AI-generated forgeries that leave no visible traces of manipulation. The shift towards image-structure analysis, which examines artefacts introduced by generation algorithms rather than the content of the document itself, represents a technical evolution that mirrors the approach taken in other deepfake detection contexts, such as synthetic video and audio detection, applied to the specific requirements of regulatory identity verification.
