Dubai has announced a new artificial intelligence model designed to detect manipulated videos and help counter the growing threat of deepfakes.
The Dubai Electronic Security Centre (DESC) introduced SARAAB during the opening day of GISEC Global 2026 at the Dubai Exhibition Centre in Expo City.
DESC says the model can detect deepfake videos with an accuracy rate of 91%. The agency describes SARAAB as the Arab world’s first deepfake detection model developed entirely by a government organisation.
The tool aims to help public and private organisations identify digitally manipulated content. It will focus on threats such as digital impersonation and fraudulent video content.
How SARAAB Detects Deepfakes
An Emirati team at DESC developed SARAAB in six months.
The model examines complete video clips rather than checking only individual images. It produces heat maps with moving hotspots that can indicate manipulated frames.
This approach allows SARAAB to examine a video from beginning to end. It can therefore identify altered frames hidden within otherwise genuine footage.
Traditional deepfake detection tools often focus on individual images or selected frames. SARAAB instead analyses the wider video sequence to identify signs of manipulation.
Deepfake technology has become increasingly sophisticated. It can create convincing videos that imitate real people and make fabricated content appear authentic.
Dubai Plans Open-Source Release
DESC plans to release SARAAB as an open-source model on Hugging Face and other developer platforms.
The open-source release will allow cybersecurity researchers and AI developers to examine the model’s source code. Developers will also be able to study its performance and adapt it for specific applications.
DESC said the model could support cybersecurity efforts across the region. Researchers could also use it to develop customised systems for different technical environments.
The agency expects the technology to help organisations respond to evolving deepfake threats. It also hopes wider access will support efforts to rebuild trust in digital media.
SARAAB could give cybersecurity teams another tool for identifying manipulated content. Its open-source design may also encourage researchers to improve deepfake detection as the technology continues to evolve.
