Singapore’s former prime minister, Lee Hsien Loong, is known to be a careful speaker. So when videos of him giving tips on investment opportunities and discussing foreign affairs made the rounds on social media platforms, they were quickly labeled as deepfakes, and taken down. They did not appear to have misled viewers, yet authorities are considering a ban on such videos ahead of a general election.
“On top of mimicking my voice and layering the fake audio over actual footage of me … scammers even synced my mouth movements with the audio,” Lee wrote in a Facebook post about one such video. “This is extremely worrying: people watching the video may be fooled into thinking that I really said those words.”
Few Singaporeans appear to have been fooled by the deepfakes of Lee, who was prime minister for two decades and is now a senior minister. Nevertheless, the deepfake showing him discussing U.S.-China relations was “dangerous and potentially harmful to our national interests,” the 72-year-old said in the Facebook post. “The motive of these deepfakes may not be commercial, but their malicious intent is clear.”
Worldwide, deepfake videos of politicians have surged in recent months as generative artificial intelligence tools become more accessible. In Singapore, deepfake videos of Prime Minister Lawrence Wong and President Tharman Shanmugaratnam were quickly taken down. While Singaporeans have a high degree of trust in their government, many think that deepfakes undermine their trust in politicians, a poll by technology firm Jumio earlier this year showed. A whopping 83% of Singaporeans surveyed said they were “worried about the potential for AI and deepfakes to influence upcoming elections.”
Compared to countries such as India, Indonesia, and the Philippines in the run-up to their elections, there aren’t as many deepfakes in Singapore. They have also not yet had the massive impact on elections that many had feared. But the recent incidents in Singapore have prompted authorities to consider a temporary ban ahead of the general election that is widely expected to be held later this year.
Singapore could follow the lead of South Korea, which introduced a 90-day ban on AI-generated content ahead of a legislative election in April, Josephine Teo, minister for digital development and information, said at a briefing in July. Under the new South Korean law, those who use deepfakes in relation to the election could be sentenced to up to seven years in prison or fined up to 50 million won ($36,000). The Philippine election commission has called for a similar ban ahead of a senate election in 2025, after several deepfake videos of President Ferdinand Marcos Jr. surfaced.
“In our context, the election period is very short. We have to find something that is suitable in our context,” Teo said.
Singapore has several laws to tackle misinformation that can also be applied to AI-generated content, the minister said. These include the Protection from Online Falsehoods and Manipulation Act (POFMA), or the so-called fake news law. Introduced in 2019, it permits a minister to declare that information posted online is “false,” and order the publisher or creator to correct or remove it. The controversial law has been invoked more than 160 times, and is so well known that it is used as a verb in the country. But it has been criticized for being wielded too often against critics of the government, including during the 2020 general election, when ministers issued several notices to opposition politicians and parties.
Singapore also has an Online Criminal Harms Act (OCHA), which empowers authorities to order service providers to act against suspicious online activities, including deepfakes.
With so many existing laws, activist Kirsten Han, who has received two POFMA orders, is wary of yet another law, she told Rest of World. “We cannot assume that the solution to any problem is to arm the state with more laws, more power, and more discretion,” said Han, a former columnist for Rest of World. Instead, allowing people to “engage in open debate, critically evaluate sources, and learn how to navigate conflict together” is preferable to having the government “regulate and arbitrate content for us,” she said.
With a tight grip over the media, Singapore ranks a lowly 129 on the Press Freedom Index. In a speech in May, Teo dismissed the opposition to the regulations as “distorted arguments about their chilling effect on free speech … or insinuations that they are instruments for silencing critics.”
Despite its discomfort with deepfakes, Singapore was an early adopter of AI: The country released its first national AI strategy in 2019, and uses generative AI tools in its courts, schools, businesses, and government agencies. Authorities have emphasized responsible use of AI “for public good.” Nevertheless, Singaporeans are among the biggest targets of investment scams and romance scams that use AI-generated content. Ads featuring deepfake images and videos of politicians and celebrities are common, and several lawmakers were targeted in an extortion scam earlier this year that used digitally manipulated images. Authorities have criticized big tech firms for not doing enough to protect users.
If a deepfake ban were to be introduced, it is unclear how effective it would be, according to Chong Ja Ian, an associate professor of political science at the National University of Singapore. The deepfakes could be created overseas, which would put them outside the reach of Singapore’s jurisdiction, he told Rest of World. Also, while deepfakes could potentially affect the outcome of close electoral races, in Singapore, “general elections involve … a single political party that is clearly dominant in resources and reach,” Chong said.
The ruling People’s Action Party has held power since the country’s founding in 1965, receiving more than 60% of the vote in every general election. With a population of about 5.5 million, Singapore is a “small, highly regulated state that generally has stronger domestic controls” than other countries grappling with deepfakes, said Chong.
Singapore has acted to blunt the impact of new technologies and platforms on previous elections: During the 2006 general election, it banned political podcasts and streaming video feeds during campaigning. Lawmakers initially derided social media as lacking in credibility; now, ministers regularly appear on podcasts, and are prolific users of Facebook, TikTok, and Instagram.
Any regulations in Singapore on deepfakes that are “overly stringent” may suppress legitimate discourse and dissent, Munira Mustaffa, executive director of security consultancy Chasseur Group, told Rest of World. “Addressing the root cause of misinformation requires comprehensive media literacy education, and critical thinking skills to discern credible sources,” she said.
“We cannot assume that the solution to any problem is to arm the state with more laws, more power, and more discretion.”
Singapore’s schools and libraries have launched efforts to educate the public about misinformation, and to fact-check content that may be AI-generated. The efforts seem to be paying off: Some 60% of Singaporeans surveyed by Jumio said they were “confident” they could easily spot a deepfake of a political figure or celebrity, compared to just a third in the U.K. and 37% in the U.S.
On Lee’s Facebook post about the deepfake video of him, several followers commented that they had not been fooled. “My first reaction to the video was that it was fake,” one user said. “Knowing how charismatic you are … you could not possibly use such vulgar words.”
