An early wave of deepfake detection software developed in Singapore has been released to the market, as the country prepares to deal with the hazards of AI-generated misinformation ahead of the general election.
Singapore's ST Engineering has released its Einstein.AI deepfake detection tool for corporations during its annual InnoTech Conference on September 3 at Marina Bay Sands. This development marks a significant milestone in the fight against misinformation, especially as deepfakes become more sophisticated and challenging to identify. The tool's release comes at a crucial time, with general elections on the horizon, where the integrity of information is paramount.
When used by banks, investment platforms, media businesses, or other clients, the technology detects irregularities in uploaded videos, such as strange brow or lip movements, as well as audio frequencies that are likely to be computer-generated, according to Mr Tan Boon Leong, a vice-president at ST Engineering. The precision of Einstein.AI in identifying these subtle anomalies is attributed to advanced machine learning algorithms that have been trained on vast datasets of both authentic and manipulated media. This ensures that the tool remains effective even as deepfake technology evolves.
"These are increasingly difficult for people to identify with their own eyes," said Mr Tan, who worked on Einstein.AI. He emphasized the importance of equipping organizations with tools that can autonomously verify content, reducing the reliance on human judgment, which can often be fallible.
After scanning, the AI provides a report that contains a deepfake likelihood score, a transcript and synopsis of the video's audio stream, and a list of associated news articles located online, allowing social media users and investors to determine the content's reliability. This comprehensive report not only aids in immediate decision-making but also serves as a valuable resource for ongoing media literacy education, helping users better understand the nuances of digital content verification.
Desmond Tan, Senior Minister of State for the Prime Minister's Office, stated in his opening presentation at the tech conference that tools such as Einstein.AI are a key step toward combating deepfakes, which increased by around 500 percent in 2023 compared to the previous year. He highlighted the government's commitment to fostering a secure digital environment, which includes supporting innovations that can preemptively address emerging threats.
Deepfakes also represent a threat to organizations, according to Mr Tan, who cited an incident in Hong Kong in which executives handed approximately HK$200 million (S$34 million) to fraudsters who used AI to impersonate their chief financial officer in order to accept a transaction.
Mr Tan stated that scammers are now utilizing advanced AI techniques to make deepfake voice recordings and movies that can fool even the most trusted family and friends. This underscores the urgent need for widespread adoption of detection tools, not just within corporations but also at the individual level, to safeguard against potential exploitation.
"With just a few photos from social media and a short voice clip—sometimes as brief as seconds—scammers can create frighteningly realistic deepfakes."
Authorities around the world are scrambling to reign in deepfakes, amid fears that fraudsters may use them to disrupt elections in numerous nations, including Singapore, in 2024. The international community is increasingly recognizing the cross-border nature of digital threats, prompting collaborative efforts to develop standardized protocols for deepfake detection and response.
India saw an epidemic of political deepfake content on social media during its elections earlier in 2024, and fake films of presidential hopefuls Kamala Harris and Donald Trump continue to circulate on social media as they campaign for the presidency.
In 2023, fake recordings of Senior Minister Lee Hsien Loong commenting on global issues and Prime Minister Lawrence Wong pushing investment initiatives appeared online. Singapore is expected to conduct its general elections in November 2025.
Concerns about deepfakes have motivated a significant investment in detecting skills here.
The Ministry of Digital Development and Information announced in May a $50 million five-year investment in the new Centre for Advanced Technologies in Online Safety (Catos), which will focus on developing technologies to detect misinformation and deepfakes. This initiative is part of a broader strategy to enhance national cybersecurity infrastructure, ensuring that Singapore remains resilient in the face of evolving digital threats.
Separately, in late 2023, developers at Singapore's Agency for Science, Technology, and Research (A*Star) announced the development of The Sleuth, a deepfake detection algorithm that can detect synthetic content on media platforms, messaging applications, and web browsers.
In April, the Home Team Science and Technology Agency demonstrated a similar detecting technology dubbed AlchemiX, which can match recordings of a suspected deepfake video to recordings of the speaker's true voice. These technological advancements reflect a growing recognition of the need for robust, scalable solutions that can be integrated across various digital platforms to ensure comprehensive protection against deepfake threats.
