Detect Deepfakesby Resemble AI
Deepfake case study · Multi-modal

North Korean hackers clone CEOs' faces and voices for

North Korean cybercriminals are using AI to clone executive faces and voices for sophisticated video call impersonation scams targeting global business.

Reported date
Nov 18, 2026
Target
founders and high-ranking business executives

The exact incident date was not established. This entry is dated by its source report.

Updated Nov 19, 2026 · 1 min read

North Korean cybercriminals linked to the Lazarus group's BlueNoroff cell are leveraging generative AI to clone the faces and voices of founders and high-ranking executives. This campaign, which has targeted 677 individuals across 58 countries, uses deepfakes to facilitate financial theft during fake Zoom and Microsoft Teams video calls.

What happened

Investigators from Arctic Wolf Labs gained access to an internal repository belonging to the attackers, revealing a sophisticated operation that targets the tech, Web3, and venture capital sectors. The attackers extract footage from YouTube interviews and conference talks to train AI models, utilizing commercial tools such as GPT-4o, Grok Imagine, Google Generative AI, and Wondershare AIGC. Forensic analysis of the attackers' archive identified 349 videos marked as likely modified, confirming the use of deepfakes that are indistinguishable from real webcam footage.

The group employs a continuous feedback loop to refine their deceptions. During fake meetings, attackers simulate technical failures to trick victims into revealing their own webcam feeds, which are then recorded and repurposed to target the victim's professional contacts. Beyond video calls, the campaign utilizes 98 malicious domains designed to mimic financial documents, security audits, and software updates for tools like Chrome’s PDF viewer and NovaStream.

Despite the advanced use of AI as a production multiplier, researchers note that the intrusion mechanism remains consistent: the attack relies on the target manually executing malicious code or installing patches. The operation is designed to scale through automation, but security experts emphasize that the chain of attack is broken if an executive verifies urgent requests through alternative, trusted communication channels.

Sources