Deepfake Detection for KYC: The Fake North Korean Military IDs

North Korean hackers used ChatGPT and GPT-4o to create deepfake military identity images. State-backed operators, using a consumer AI product, to manufacture fake IDs. The barrier to forged identity documents has never been lower.

Why this terrifies compliance teams

Fake IDs are the skeleton key of financial crime. Open bank accounts, pass KYC checks, move funds, all under a synthetic identity backed by convincing documents. When the forger is a state actor with AI image generation, the fakes are not crude Photoshop jobs. They are clean, consistent and tailored to pass automated checks.

The North Korean angle adds a sanctions dimension. These are not just fraudsters; they are operators funding a weapons programme through cybercrime. Every fake ID that passes a compliance check is a potential sanctions violation for the institution that accepted it. Regulators do not accept "the AI fooled us" as a defence. The irony is sharp: a consumer AI product, built with safety filters, being used by a sanctioned regime to forge documents. Filters catch the obvious prompts. Determined state actors do not use obvious prompts.

Deepfake detection for kyc is the control built for this threat. To detect deepfake videos in onboarding and flag synthetic identity documents, you need forensic checks at the point of verification, not after the account is open. A deepfake detection api for kyc screens the document and the selfie video together, catching the forgeries that fool human reviewers and basic automation alike.