Dubai is preparing SARAAB, an open-source AI model for deepfake detection that analyses facial features and motion in videos. The Dubai Electronic Security Centre (DESC) says the model can identify hidden synthetic layers in footage, including cases where only part of a video has been manipulated.
SARAAB is expected to launch by the end of 2026. DESC announced it at GISEC Global 2026, which runs from 16–18 September at Dubai Exhibition Centre in Expo City Dubai. The announcement was first reported by Khaleej Times, while Gulf News reported that the model was developed by an Emirati DESC team.
How deepfake detection works in SARAAB
The model analyses facial features through a heat map, looking for synthetic layers applied to a video. Hassan Majid Alkhazraji, Senior Artificial Intelligence Executive at DESC, told Khaleej Times that SARAAB would analyse a full two-minute video even if the suspected manipulation appeared only in the final 30 seconds.
DESC claims an accuracy rate of 91 per cent, although the announcement does not publish the test dataset, evaluation method or comparison conditions behind that figure. It is therefore a claim about the model rather than an independently verified benchmark.
SARAAB will be open source
DESC says SARAAB will be made available through developer platforms, allowing AI engineers, cybersecurity specialists and researchers to test and improve it. The Khaleej Times report names Hugging Face as the planned platform, but a public model page or release date was not available in the sources reviewed for this draft.
That open-source approach could make the model useful to organisations checking suspicious footage, but detection results still need context. A score from one model is not proof that a video is authentic or manipulated, particularly as synthetic video techniques change.
Illustrative image: a UAE cybersecurity technology partnership at GISEC Global; this image does not show SARAAB.
DESC also unveiled a Dubai Cyber Skills Framework and an ISR Auditor Certification Programme at GISEC Global 2026. The initiatives place SARAAB inside a wider push to build local cybersecurity tools and skills, rather than treating deepfake detection as a standalone software release.
Dubai has also been expanding its deepfake verification tools, while the UAE Cybersecurity Council’s AI warnings underline the wider security questions around synthetic media.
What is SARAAB?
SARAAB is an AI model being developed by Dubai’s Electronic Security Centre to detect synthetic layers in deepfake videos.
When will SARAAB launch?
DESC says SARAAB is expected to launch by the end of 2026, but a precise release date was not provided.
Will SARAAB be available to the public?
DESC says the model will be available through developer platforms so engineers and researchers can test and improve it. A public model page was not available in the sources reviewed.
