AI Against Disinformation, Defamation, and Hatred

The Indonesian government is currently working vigorously to eradicate the disinformation, defamation and hatred caused by the misuse of social media and the unethical development of Artificial Intelligence (AI). We are facing an AI technology that was born as an extension of human reason (Good AI), but has quietly transformed into the tidiest machine for multiplying lies ever created (Evil AI).

It is no longer a question of whether AI is dangerous or beneficial, because the answer lies bare before our eyes: both are true at once. Our shared homework is how to ensure this dual power heals the public sphere more often than it wounds it. In other words, how we can use Good AI to eradicate Evil AI.

Consider the data. Between October 2024 and October 2025 alone, Mafindo recorded 1,593 hoaxes, most of which were born from the womb of AI, ranging from deepfakes to personalised scam schemes that appear to know their victims personally. These are no longer crude chain messages that are easily exposed as false. Videos, voices, websites — all can now be forged with a precision that dulls our eyes and ears, which have long been our mainstay for telling the real from the fake.

And those figures pale in comparison with recent developments. Deputy Minister of Communications and Digital Affairs Nezar Patria, in September 2025, cited Sensity AI data: deepfake content has surged 550 percent over the past five years, and nine out of ten instances are used for harmful purposes. Early 2026 looks even more alarming. A cybersecurity workshop in Yogyakarta presented data showing a 1,400 percent spike in deepfake attacks within a year, accompanied by success rates of 54 to 60 percent for AI-based phishing attacks. Septiaji Eko Nugroho, Chair of the Mafindo Presidium, put it this way: deepfakes in Indonesia are now most commonly used to defraud people, whilst simultaneously steering public opinion on the political stage.

When Algorithms Stoke Hatred

The problem does not stop at emptied wallets. The Chairman of Bawaslu, Rahmat Bagja, ahead of preparations for the 2029 General Election, warned that AI enables anyone to produce content that is nearly impossible to distinguish from reality, worsened by bots, coordinated account networks and micro-targeting that hunts specific voter groups like guided missiles. More heartbreaking still, the same data shows that at least 11 percent of young women aged 15 to 29 have experienced online gender-based violence since their teens, mostly through deepfakes and visual manipulation.

This is no longer abstract discourse on the ethics of technology; it is a real wound borne by real people. Herein lies the true danger, and perhaps here the proverb ‘still waters run deep’ fits best. Sensationalist hoax content more easily provokes emotion, and technically neutral social media algorithms prioritise it precisely because it is considered ‘engaging’ by many. Echo chambers form, locking netizens inside bubbles that align with their own beliefs. Defamation and hate speech flourish, not only from easily flagged anonymous accounts, but also by sneaking quietly into information that looks convincing, yet is murky at its core.

The Counter-Weapon: AI as Detector

But this story must not end as mere complaint. The same knife can slice bread or cut a hand — it depends on who wields it. Mafindo’s 2025 study shows that transformer-based language models are capable of classifying hoax news in Indonesian with accuracy above 90 percent.

Meanwhile, research published in the June 2026 edition of the Khatulistiwa Informatika journal has gone a step further, developing an autonomous AI-based hoax detection system with 89.3 percent accuracy that can even present the evidence behind its decisions, rather than simply declaring ‘this is a hoax’ without justification.

Civil society is not standing idle on the sidelines either. Mafindo runs TurnBackHoax.id as a fact-checking data repository, and has partnered with independent developers to launch Ceksumber, an AI-based hoax-checking tool accessible through WhatsApp messages, as easy as greeting an old friend. This collaboration has also extended to Google through Cekfakta.com, uniting Mafindo, the Alliance of Independent Journalists and the Indonesian Cyber Media Association in a single verification front.

All of this affirms what has perhaps become an open secret: AI is a double-edged knife, and the direction of its sharpness depends on who grips the handle. The very same artificial intelligence used by xAI to train Grok in real time from the X timeline — so much so that it was temporarily blocked by the Ministry of Communications and Digital Affairs earlier this year — could in fact be engineered to safeguard human dignity, provided it is designed from the outset with ethical alignment that is not half-hearted.

An Idea: A Digital Authenticity Chain and an Automated Risk Score

There is one layer of defence that is rarely noticed: blockchain as the guardian of the provenance of information. If generative AI is a machine that manufactures doubt, distributed ledger technology works on the reverse logic: permanently recording the origin and modification history of data, immune to unilateral tampering by anyone.

From our experience designing blockchain systems for resolving electronic information disputes for the Central Information Commission, plus blockchain-based traceability in the global halal supply chain with the Halal Product Assurance Organising Agency, we believe there is one step that has never been proposed in full in this country: uniting the recording of content provenance, AI-based risk assessment, and the legal dispute mechanism within a single interlocking architecture.

The concept is not actually complicated; it simply has never been executed at national scale. Imagine this: every piece of generative AI content must be stamped with a cryptographic signature and time-stamped on a national distributed ledger before it may circulate on digital platforms operating in Indonesia. Not merely a water-mark sticker that can be removed with a single click, but a trace embedded at the protocol level, impossible to forge undetected.

As soon as content is reported by netizens, an independent AI assessment model immediately calculates its risk score, measuring the potential for defamation or hate speech it contains, then forwards high-scoring cases to the information dispute resolution channel that we are currently building for the Central Information Commission. In this way, provenance verification, danger detection and law enforcement no longer run separately like three arrows flying off course in different directions, but as one chain of trust moving in unison. This is the bridge connecting intelligent computing with dispute resolution science — two fields that have long run as parallel roads, rarely intersecting in Indonesia’s digital policy.

Regulation Is Not Mere Formality

Technical strength alone will not suffice without a firm legal umbrella. the Electronic Information and Transactions Law and the Personal Data Protection Law serve as the foundation for demanding accountability from Electronic System Operators, including global platforms reaping huge profits in Indonesia. The government has gone further with Government Regulation Number 17 of 2025 on the Governance of Electronic System Operation for Child Protection, better known as PP TUNAS, born directly from concern over the impact of deepfakes on vulnerable groups. Demanding compliance from global platforms does not mean being anti-innovation. This is about safeguarding the nation’s information sovereignty so it does not become a sacrifice for the economic interests of foreign technology players who sometimes act as if they stand above the law of any country.

Literacy and Conscience as the Last Firewall

However, regulation and detection algorithms are only half the journey, not the finishing line. Research by Alamsyah and Lestari of Universitas Swadaya Gunung Jati, published in the Sibatik journal in 2026, found something rather piercing: many social media users fail to evaluate visual information critically, not out of malice, but simply because they were never equipped with that ability from the start. The burden of verification must therefore not be laid entirely on individuals. Hate speech and defamation do not always arrive wrapped in easily recognisable hoaxes; sometimes they sneak inside information whose facts are true, but wrapped in rotten intent.

This is where the importance of a collaborative verification ecosystem lies — one built through gotong royong between government, platforms, academics, media and volunteer communities like Mafindo, which now reaches 44 regions across Indonesia. However sophisticated the detection technology, it is all in vain if society remains easily provoked by sensational headlines and too lazy to check the truth before their fingers hit the share button.

Towards a Digital Sovereignty with Conscience

We are not talking about stemming the tide of innovation, for that is clearly impossible — and truthfully, there is no need. What is needed is the courage to navigate its risks with a compass of humanity that must not deviate in the slightest, like a ship’s captain remaining calm amid great waves. Indonesia already has the capital for this, from hoax-detection initiatives developed by domestic researchers to cross-sector collaborations such as Cekfakta.com.

By way of illustration, BRAIN IPB University itself has pioneered the application of artificial intelligence to a public information dispute resolution system, cybersecurity resilience testing for the national health service in partnership with BPJS Kesehatan, and a blockchain-based traceability model for safeguarding food data integrity. This is concrete proof that digital sovereignty can be knitted from domestic laboratories and campuses, not merely a slogan on the seminar stage.

The task ahead is to ensure that Indonesia’s young generation stops being spectators on the side-lines whilst foreign algorithms fight in the centre of the field. They must dare to step forward as designers of verification systems and architects of the nation’s own digital ethics. For in the end, the battle against disinformation, defamation and hatred is not about who owns the most sophisticated machine, but about who succeeds in entrusting the spirit of humanity into every line of their code, however small that line may appear on screen.