KUALA LUMPUR: The production and spread of deepfake artificial intelligence (AI) videos on social media platforms, particularly those featuring political figures and addressing sensitive issues, pose a significant threat to national security.
Universiti Sains Malaysia Cyber Security Research Centre director Prof Dr Selvakumar Manickam said that in extreme scenarios, deepfake AI videos could be used to simulate attacks or create 'false flag' operations that could potentially trigger conflict or war in a country.
"The dissemination of deepfake videos, especially those involving political figures or (playing up) sensitive issues, could be considered a threat to national security.
"The ease of creating and distributing convincing deepfake (AI videos) will further increase the threat (to national security), making it (as) a powerful tool for enemies who want to destabilise a country," he told Bernama recently.
He said this in response to the recent emergence of AI deepfake videos involving two leaders, namely United States President Donald Trump and Prime Minister Datuk Seri Anwar Ibrahim, and its major implications for national and international security.
The AI deepfake video that went viral on social media platforms shows Trump and Anwar engaged in conversation, discussing geopolitical issues and bilateral diplomatic relations while at the same time promoting an investment scheme that supposedly offers multiple profits.
The AI deepfake video involving images of the two leaders was first detected on the X and TikTok platforms before spreading to other social media platforms such as Facebook and WhatsApp.
To date, neither the US embassy in Malaysia nor the Prime Minister's Office has made any statements about the spread of the AI deepfake video showing their leaders chatting.
However, if the production and distribution of such AI deepfake videos is not curbed, it could affect political stability and undermine public trust in the country's leadership, Selvakumar says.
He does not rule out the possibility that AI deepfake videos have the potential to become a powerful tool for vested interests to exploit political sensitivities and social dynamics in Malaysia, thereby creating conflict or manipulating public opinion.
According to him, there have been previous cases where AI deepfake videos have had a major impact on world leaders and major events that took place in a country.
He cited a case in Indonesia, where an AI deepfake video was spread to depict a former president (now deceased) encouraging the people to go out and vote.
"In India, AI deepfake recordings were used to spread false statements attributed to a political figure, while in Slovakia, AI deepfake audio recordings with a voice mimic of a liberal candidate were distributed before the election, which is believed to have affected the outcome of the country's election," he said.
Selvakumar said that the AI deepfake technology used in the videos involving the Malaysian and the US leaders showed advances in AI technology to the point of being able to imitate lip movements, voice intonation and facial expressions with a high level of accuracy.
AI deepfake video detection, he said, is relatively easy for those who are trained and can be identified through technical aspects such as imperfection of mouth movements with audio, inconsistencies in lighting and shadows on faces and objects.
"However, for the general public who lack technical expertise, a different approach is needed to avoid being fooled by deepfake videos," said Selvakumar, advising users to check their news sources carefully.
Previously, several AI deepfake videos involving top management of Petronas, Tesla Inc CEO Elon Musk and Seputeh MP Teresa Kok have also gone viral on the country's social media platforms. - Bernama
