باحثون سعوديون يطورون نظام ذكاء اصطناعي يكشف الأخبار المضللة بدقة توصل 98.23%Saudi researchers build an AI system that spots misleading news with up to 98.23% accuracy

Researchers from the University of Tabuk and King Saud University, working with a researcher from Arizona State University in the US, have built an AI system that detects fake and misleading news during wars and crises, Okaz reported. The study was published in Scientific Reports, a Springer Nature journal, on 16 October 2026.

The idea is that misinformation in a crisis is not always an outright lie. Sometimes it combines text, an image and an emotional charge, each of which looks fine on its own, while the deception comes from how they are put together. So the system analyses a news item along three paths: the text with a multilingual language model, the image with computer vision, and the emotions and emotional tone that can be exploited to build misleading stories.

The researchers tested the system on four well-known misinformation datasets. It scored 94.66% accuracy on Fakedet, 98.23% on Ukraine-GitHub, which is linked to the Ukraine crisis, 98.12% on Weibo and 97.98% on MediaEval. According to the study, combining the three paths consistently beat a set of recent methods the researchers compared it with.

But the results also show the system still needs to be tested on Arabic content and local dialects, on news as it spreads the moment events happen, and on images and videos created or edited with AI.