“A picture is worth a thousand words”, but in the age of Artificial Intelligence (AI), the picture may not be worth anything at all, or worse, it may be worth a thousand lies. A deepfake is an AI-generated image, video or audio that convincingly depicts people doing or saying things they never did. For a high-trust, hyper-connected society, social cohesion is both a point of national pride and political necessity. However, the rise of malicious deepfakes threatens this very social fabric that Singapore spent decades building.
The term “Deepfake” first came about in 2017 by an anonymous Reddit User who used open-source face-swapping technology to create non-consensual pornographic content. Today, with advances in AI, tools like OpenAI’s Sora can produce hyper-realistic video from simple text. The societal costs compound quickly as AI-enabled impersonation has increased the effectiveness of financial scams, and victims of deepfake sexual abuse suffer lasting psychological trauma. Furthermore, in a compact and multiracial society, a single viral deepfake engineered to inflame religious or racial sentiment could undo decades of trust.
Current Issue: When Reality is Distorted
Recently, in June 2025, a fabricated video of Senior Minister Lee Hsien Loong making inflammatory remarks about China circulated on an encrypted messaging platform before it was debunked. It appeared sophisticated enough to be credible without context; so it had the potential to trigger a diplomatic fallout.
Months later, another incident took place at the Singapore Sports School where students became victims of AI-generated non-consensual intimate images produced by automated bots and distributed through a close online group. Despite law enforcement intervening, copies continue to resurface. Victims withdrew from digital life, traumatised by the event. The harm was cumulative, persistent and irreversible
These incidents signalled an early sign of a structural problem in Singapore’s regulatory posture. Deepfakes can be created in minutes and shared globally, while verification and platform enforcement operate on a timescale of hours to days. By the time any institution responded, the content had already been viewed, saved and redistributed across multiple platforms and jurisdictions. A takedown order does not unsee an image and the damage has already been done.
The current policies and measures are a reactive stance and become ineffective when the deepfake is generated in another jurisdiction.
Existing Policies and Measures
The Fight Against Perpetrators
The Protection From Online Falsehoods and Manipulation Act (POFMA) was enacted in 2019, and enabled Singapore to take down or block false statements which undermine public interest. This effectively allowed deepfakes spreading false information to be taken down directly. However, this cannot be applied to deepfakes which do not make factual claims yet still cause potential societal harm.
In addition, POFMA worked hand in hand with the Maintenance of Religious Harmony Act (MRHA) which was enacted in 1990 with amendments made in 2019 to better target digital threats.Similar to POFMA, the MRHA allowed for the takedown of online content, but was targeted at content that created religious enmity.
This is supported by the Online Criminal Harms Act (OCHA) enacted in the early 2020s, the Protection from Harassment Act (POHA) in 2014 and the penal code itself.Beyond the prevention of societal harm, these acts prevent the usage of deepfakes in criminal activities. The OCHA enabled authorities to disable online accounts linked to these activities. The penal code criminalises the use of deepfakes in scams, hate speech, sexual offences and more. While POHA allowed for protection of victims through protection orders, correction orders and monetary compensation while also punishing the perpetrator. However, these acts are all reactive in nature, made to take action after the criminal act occurred, but failed to tackle the root issue of the usage of deepfakes themselves.
Protecting The Innocent
The final piece was the Online Safety (Relief and Accountability) Bill (OSRA) and the Online Safety Commission (OSC) which would take effect on 29 June 2026.In contrast to the other acts which focused on the perpetrator and punishing them, the OSRA focuses on support rendered to victims of deepfakes, with the OSC being the governing body to support this. It allows victims to file reports directly to a government body, instead of reporting only to social media sites, along with providing victims with legal advice, disclosing the perpetrators information and legal clinics to bring official claims against them.
The Dilemma for Policymakers
Singapore’s dilemma was that every policy option carried trade-offs. Enforcement agencies needed speed and deterrence to close the gap between the rapid spread of deepfakes and the slower pace of investigation. Legal authorities, however, had to ensure that any preventive measure was proportionate, clearly scoped and did not become an overly broad restriction on satire, art, commentary or legitimate expression.
Economic agencies also had to consider Singapore’s reputation as a trusted AI and technology hub. Clear rules against harmful misuse could strengthen trust, but broad or unpredictable controls could deter innovation. Civil society faced a similar tension. Youth protection advocates may support stronger intervention against deepfake abuse, especially where minors were targeted, while digital rights advocates may worry that identity verification or monitoring obligations could weaken privacy.
Existing laws already gave Singapore tools to address parts of the problem, including online falsehoods, criminal online harms, harassment and threats to religious harmony.Yet these tools were often strongest after harm had already surfaced. For victims, institutions and the wider public, the damage caused by a convincing fake could begin before a formal remedy was available.
Conclusion
The most viable response was therefore a targeted hybrid model. Singapore should avoid regulating deepfake technology as a whole. Instead, it should focus on clearly defined high-risk harms, such as political impersonation, intimate-image abuse, scams and content that could inflame communal tensions. Faster reporting pathways, stronger platform accountability and victim-support mechanisms would help reduce harm without closing off legitimate uses of synthetic media.
Ultimately, deepfakes tested whether Singapore could preserve trust without freezing innovation, and act quickly without overreaching. In an age where seeing was no longer believing, trust had to rest on institutions, norms and safeguards that helped society decide what was true.