Refusing to POFMA AI disinformation targeting PM Wong exposes what the law is really for

Before examining what the government did not do in response to the AI-generated disinformation campaign targeting Prime Minister Lawrence Wong, it is worth dwelling on what Parliament was told when POFMA was passed in May 2019. Not paraphrased. Not summarised. What was actually said.

Then-Law Minister K Shanmugam told Parliament during the Second Reading of the Protection from Online Falsehoods and Manipulation Bill:

"Singapore is a specific and vulnerable target for some very precise reasons. It has military superiority in this region. That superiority in conventional military forces means it will be futile to start a war with Singapore. And, therefore, militarily weaker countries will then focus on other means to weaken Singapore, sap our will from inside, create deep internal divisions and keep us in a permanent state of internal dissension. The evidence is that this is already happening."

On who the law was primarily aimed at, Shanmugam was equally direct in his reply speech:

"It does not make sense to go and find and look for thousands of people, and individually go to each one of them. It makes sense if there is one creator of falsehoods, deliberately creating and sitting somewhere and pushing out falsehoods, that person can be dealt with."

Then-Minister for Communications and Information S Iswaran, speaking in the same debate, was unambiguous about the Bill's primary focus:

"The primary focus of this Bill is not individual per se; it is actually the larger tech platforms."

Then-Senior Minister of State for Law Edwin Tong, who now serves as Law Minister, told Parliament in the same sitting:

"Large majority of the toolkit is therefore designed for platforms, not individual publishers."

These are not ambiguous statements. They are not open to creative interpretation. They describe, with precision, a law designed to confront coordinated, platform-amplified, foreign-linked disinformation operations — not individual Singaporeans posting contested views on social media.

Now consider what has just unfolded. Nearly 300 AI-generated Chinese-language videos. Deepfake avatars. Channels created in coordinated 20-minute windows. Millions of views. An operation assessed by academics as likely state-backed or well-funded, targeting Singapore's Prime Minister with fabricated narratives designed to destabilise public confidence in his leadership. Creators sitting somewhere, deliberately pushing out falsehoods, at scale, on platforms, targeting Singapore.

If that is not the threat POFMA was built for, one must ask what exactly it was built for.

The answer, judging by six years of the law's application, appears rather different from what Parliament was told.

A convenient justification

When Member of Parliament Fadli Fawzi of the Workers' Party asked Minister for Digital Development and Information Josephine Teo in Parliament why no POFMA correction direction had been issued against the AI-generated videos targeting PM Wong, her response was disarmingly simple: YouTube had already removed most of the accounts, so there was no need.

It sounds reasonable on its face. But it does not hold up to scrutiny.

Lianhe Zaobao was the first to report on the campaign. Channel NewsAsia (CNA) followed with a deeper investigation days later. Between those two reports, the operation continued. Videos and channels resurfaced rapidly after removal, as Josephine Teo herself acknowledged in the same parliamentary answer.

By the time CNA published its findings, videos and accounts were still accessible. The premise that content had been substantially removed before action was considered does not align with the timeline of events on the ground.

More fundamentally, the government's justification rests on a standard it does not apply consistently.

The case of Jay Ish'haq Rajoo

In September 2025, a POFMA correction direction was issued to TikTok user Jay Ish'haq Rajoo over a video he had posted on 26 August 2025 concerning a Chinese community leadership course.

Jay had alleged in the video that National Development Minister Chee Hong Tat had announced government plans to provide money and resources to attract mainland Chinese to Singapore and groom them for leadership roles, and that government resources were being used solely to nurture leaders from the Chinese race.

The Ministry of Culture, Community and Youth (MCCY) said these claims were false and could cause serious misunderstandings and erode public confidence in the government's commitment to serve Singaporeans regardless of race.

The direction was issued on 7 September 2025, on the instruction of Acting Minister for Culture, Community and Youth David Neo — 12 days after the video was posted.

By that date, the original video was no longer accessible. According to Jay, he had removed it within a day of uploading it. The video was gone. The alleged falsehood had ceased to circulate. And yet a formal POFMA direction was issued anyway, requiring Jay to publish a new post carrying a correction notice and a link to the government's clarification.

A private citizen posted a video, removed it within 24 hours, and still received a formal legal direction 12 days later. Meanwhile, a coordinated foreign-linked operation producing hundreds of AI-fabricated videos, accumulating millions of views, targeting a sitting prime minister with fabricated narratives about his removal from office — that did not meet the threshold for POFMA action, because YouTube had removed some of the accounts.

If the removal of content is sufficient reason not to invoke POFMA, why was Jay subjected to a correction direction after his video was already gone? No official explanation bridges these two outcomes. The standards of the POFMA Office appear, at best, undefined. At worst, they appear to be applied selectively.

It is also worth noting that the September 2025 direction was not Jay's first encounter with POFMA. In August 2023 alone, he received three separate correction directions from three different ministers — Chan Chun Sing, Tan See Leng, and Desmond Lee — issued within a single week, over three separate posts on CPF, housing, and voting.

He received a conditional warning in July 2024. One private citizen, repeatedly targeted across multiple years and multiple ministers. Zero directions against a 300-video foreign-linked influence operation targeting the Prime Minister.

The communications infrastructure argument

There is a further dimension worth examining.

During the Committee of Supply debate on 2 March 2026, Josephine Teo told Parliament that Singapore's public service media entities — The Straits Times, CNA, and Lianhe Zaobao among them — reach over 90 per cent of Singaporeans and are more trusted than reputable international outlets. She described them as indispensable to countering misinformation.

The government's financial commitment to these outlets is not modest. Approximately S$380 million flows annually to Mediacorp, which operates CNA and several broadcast channels.

SPH Media Trust, publisher of The Straits Times and Lianhe Zaobao, receives up to S$180 million annually as part of a commitment of up to S$900 million over five years. That is over half a billion dollars a year directed at outlets the government itself describes as the nation's frontline defence against misinformation.

Any minister can issue a statement, hold a press conference, or brief a journalist within the hour. The government's rebuttal of the AI videos — calling them "wholly fabricated and outlandish" — was duly carried by precisely these outlets, reaching precisely those 90 per cent of Singaporeans Teo cited.

So what does a POFMA correction direction add that this formidable communications infrastructure cannot achieve on its own?

The answer is enforcement, legal designation, and compelled compliance. A correction direction formally brands content as a falsehood under law, requires the original publisher to carry that designation, and creates an enforceable public record. It is not a communications tool. It is a legal one.

Which makes the choices here all the more stark. Against a private citizen whose video had already been removed, the government reached for the legal instrument. Against a foreign-linked operation producing hundreds of AI-fabricated videos targeting the Prime Minister, it issued a press statement and pointed to its well-funded media.

If the government's own communications infrastructure is sufficient to counter the most serious disinformation campaign Singapore has seen, it is difficult to understand why POFMA was ever necessary for the domestic cases it has actually been applied to.

And if that infrastructure is not sufficient — if legal designation and compelled correction genuinely matter — then the question of why the law was not used here becomes considerably harder to answer.

A law that looks inward

The pattern of POFMA's use since 2019 is instructive. According to the POFMA Office's own data, the law has been invoked in 89 cases to date, resulting in 143 Correction Directions, 39 Targeted Correction Directions, and five General Correction Directions, among other instruments.

The overwhelming majority of those 89 cases have been directed at politicians, independent media outlets, activists, and individual content creators — precisely the domestic civil society actors that critics warned about during the law's parliamentary debate.

Opposition politicians have not been spared. Former Non-Constituency Member of Parliament (NCMP) Leong Mun Wai and former NCMP Yee Jenn Jong have both been issued POFMA directions. Independent media outlets and ordinary social media users have received correction directions over posts that were, in some cases, matters of contested interpretation rather than clear fabrication.

There is a particular irony in who championed the law most loudly and whose ministerial portfolio wielded it most extensively. Shanmugam, the minister who spent hours before Parliament in 2019 warning of foreign interference, bots, and coordinated inauthentic behaviour, went on to become the law's single most prolific individual user, personally issuing 20 of the 89 total directions on record.

But the fuller picture is more striking still. During his tenure as Minister for Law and Home Affairs, the two ministries under his stewardship collectively issued 36 of those 89 directions — more than four in every ten — including directions issued by Edwin Tong as Senior Minister of State for Law and by Josephine Teo as Second Minister for Home Affairs.

Not one of those 36 directions targeted a foreign actor, a coordinated inauthentic network, or an AI-generated influence operation — the very threats Shanmugam spent hours warning Parliament about in 2019. They were aimed, without exception, at domestic targets: opposition figures, civil society voices, activists, anti-death penalty advocates, independent journalists, and media outlets.

The Workers' Party, which voted against the Bill in 2019, warned during the debate that the law vested ministers with powers that could be turned against critics and domestic political actors, while offering no equivalent protection to the public from coordinated foreign disinformation. Shanmugam dismissed such concerns as melodramatic. Six years on, the record speaks for itself.

What the speed argument actually shows

The most telling contradiction, however, lies not in who was targeted but in when.

The central justification for vesting POFMA's powers in the Executive rather than the courts was speed. Shanmugam and Tong returned to this argument repeatedly in 2019.

Falsehoods moved too fast. Courts could not act quickly enough. Breaking virality required immediate Executive action. This was not a minor supporting point — it was the government's answer to every critic who said an independent body or the judiciary should decide first.

The POFMA Office's own data comprehensively undermines the speed justification.

Of the 90 directions issued with a known gap between the original post and the direction, just one was issued on the same day. Only 15 were issued within 24 hours — and the overwhelming majority of those involved COVID-19 health misinformation during a genuine public emergency. The average gap between an offending post and a POFMA direction across all cases was 17.3 days. The median was five days.

Cases outside the COVID period tell a consistent story. A direction against The Online Citizen over the AIM software controversy took 35 days. A direction against former NCMP Yee Jenn Jong over the same matter took 32 days. A direction against Kirsten Han, Andrew Loh, The Online Citizen, and the Transformative Justice Collective over statements concerning a death row inmate took 30 days. A direction against the East Asia Forum over an article on Singapore's political scandals took 26 days.

These are not the timelines of a law being used to break virality. They are the timelines of a law being used for correction, for the record, and in some cases for legal punctuation long after any original spread had run its course. The speed argument, it turns out, was largely reserved for a public health emergency. For everything else, the government took its time — while insisting the law existed precisely because time was the one thing it could not afford to waste.

The monetisation pipeline

There is a further dimension to how correction directions function in practice that goes largely unreported. Under POFMA, an online location that accumulates three or more directions within a six-month period can be designated as a Declared Online Location — a status that prohibits the site from carrying paid content or accepting advertising revenue in Singapore. It is, in effect, a legally enforceable monetisation kill switch.

The data shows this pipeline has been deployed systematically against Singapore's independent media landscape. The Online Citizen has been designated a Declared Online Location twice — in July 2023 and again in June 2025. Gutzy Asia, which emerged as a successor publication to The Online Citizen, was itself declared in June 2024.

Between them, the same independent media operation has been subjected to three separate Declared Online Location designations across two mastheads. Kenneth Jeyaretnam's publication was declared in December 2023. The Transformative Justice Collective was declared in December 2024.

The connection to Shanmugam's ministries runs through virtually every one of these designations. The Online Citizen's first DOL in July 2023 was triggered by two correction directions both from MHA — one issued personally by Shanmugam, one by Josephine Teo as Second Minister for Home Affairs.

The second DOL, in June 2025, was triggered by two directions issued by Edwin Tong as then-Second Minister for Law — both concerning Shanmugam personally. The first related to TOC's reporting on GCB transparency, which had referenced his property sale; the second targeted TOC's reporting on the Ridout Road earthworks at his state-leased residence. A law minister's ministerial subordinate issued legal directions over articles that were, in both cases, primarily about the law minister himself. The directions then formed the basis for stripping that outlet of its ability to earn revenue in Singapore.

The Transformative Justice Collective's December 2024 DOL was preceded by five directions from MHA alone, three issued personally by Shanmugam and two by Josephine Teo. Kenneth Jeyaretnam's DOL was triggered primarily by a cluster of MinLaw directions under Shanmugam's stewardship. The pipeline from correction direction to financial designation ran, in case after case, through the same ministerial portfolio.

Every entity ever designated as a Declared Online Location under POFMA is a domestic civil society actor, independent media outlet, or opposition-linked voice. Not one foreign-operated disinformation network, not one AI-generated influence operation, not one coordinated inauthentic channel has ever been declared.

This matters because it fundamentally reframes what correction directions actually are. They are not merely factual corrections placed alongside contested content, as ministers described them in 2019. In practice, they function as the first stage of a legal process that can strip an independent publication of its ability to sustain itself financially.

The law that was sold to Parliament as a tool to break the virality of foreign disinformation has become, in application, an instrument capable of breaking the financial viability of domestic independent media — pursued, in the case of one editorial operation, across three designations and two different publication names.

The public discernment reversal

Perhaps the sharpest irony of Josephine Teo's response to Fadli lies in its central argument: that a discerning public is the first and most important line of defence against disinformation, and that the government's role is to encourage Singaporeans to verify content and consult official sources.

That is precisely what opposition members and civil society voices argued in 2019 — and it is precisely the argument the government defeated in order to pass POFMA.

During the Second Reading debate, critics of the Bill questioned why legislation was necessary when the better long-term answer was investing in media literacy and trusting citizens to evaluate information critically.

The government's response was emphatic: media literacy was necessary but not sufficient. Falsehoods exploited cognitive biases too effectively. Public discernment alone was not a meaningful defence. Executive action with speed was essential.

That argument won. Parliament passed the law. Ministers were vested with powers to act swiftly against coordinated online falsehoods in the public interest, without waiting for courts or relying on platform self-regulation.

Now, faced with the most serious and sustained disinformation campaign Singapore has seen — one targeting the Prime Minister himself, built on precisely the tools and tactics that ministers warned Parliament about — the government has reached for the very argument it once defeated. Trust the public to be discerning. Check official sources. Rely on well-funded public service media.

It is difficult to reconcile these two positions. Either public discernment was always sufficient, in which case POFMA should never have been passed. Or it is insufficient in the face of coordinated foreign disinformation at scale, in which case POFMA should have been used here. The government cannot credibly occupy both positions depending on which outcome it prefers.

The asymmetry across 89 cases is stark. The law, as applied, appears far better calibrated to protect the government from domestic criticism than to protect Singaporeans from foreign disinformation.

That is not what Parliament was told in 2019. The quotes at the top of this piece make that plain. And it is a question Singapore deserves a direct answer to.