Guarding the Sovereignty of Voters Against AI Microtargeting
Electoral democracy today is no longer merely tested by mass mobilisation in open fields through campaign rallies or vision-and-mission debates on formal stages. The real political battlefield has now shifted into a very quiet, invisible space: the cyber realm controlled by artificial intelligence (AI) algorithms.
Amid the dominance of young and first-time voters, accounting for more than 56% of the 214.35 million voters on the national register (General Election Commission/KPU RI, 2026), we face a new, systematic threat to citizens’ autonomy of thought: political microtargeting-based manipulation of public opinion. Protecting the sanctity of the vote must now contend with leaps in computing technology that move far faster than our legal adaptation. Whereas in the past algorithms operated linearly according to rigid rules written by humans, modern AI algorithms work autonomously, finding their own patterns. Machines are simply fed massive behavioural data and a single objective: to maximise users’ watch-time and engagement on the platform. Algorithms are not designed to care about factual validity, narrative proportionality, or democratic learning for citizens.
Within this feedback loop, political microtargeting finds its ammunition. Through the processing of massive digital activity data — from what users like, how long they linger on content, to their active hours — AI can map voters psychologically. Voters are no longer grouped conventionally by demographic variables such as age or region of origin, but by their emotional tendencies and psychological vulnerabilities: whether they are anxious about the economic future, angry about identity issues, or doubtful about incumbents.
Once the psychological mapping is complete, computational models design highly personalised campaign messages to influence their behaviour. Dangerously, these messages are sent as dark posts directly to each individual’s private feed. This pattern creates acute information asymmetry: the public does not know what messages their neighbours receive, and election oversight bodies lose their supervisory capacity because those messages never appear in public space.
A healthy political campaign should be a forum for public dialectic where arguments are tested transparently. Yet the exploitation of voter data degrades this essence by confining citizens within artificial echo chambers. Through collaborative filtering, users are grouped with like-minded individuals, filtering out alternative views so they never surface. In this homogeneous ecosystem, manipulative narratives are repeated en masse by dozens of synthetic sources, triggering the illusory truth effect, whereby the human brain cognitively treats repetition as absolute truth.
PERSONAL DATA PROTECTION
At this point, personal data protection is no longer merely a technical matter of information security or administrative protection. Privacy is an integral part of human rights protected by international legal instruments such as Article 12 of the 1948 Universal Declaration of Human Rights (UDHR), Article 17 of the 1966 International Covenant on Civil and Political Rights (ICCPR), and Article 28G paragraph (1) of the 1945 Constitution of the Republic of Indonesia (UUD RI).
The enactment of Law Number 27 of 2022 on Personal Data Protection (PDP Law) should be operationalised as a constitutional instrument to enforce ethical limits on data processing in elections. The massive processing and exploitation of voter data profiles for micro-political targeting without valid consent constitutes a serious violation of the data sovereignty of voter subjects.
Unfortunately, law enforcement against this cyber manipulation is hampered by a wide regulatory gap. For example, Constitutional Court Decision Number 166/PUU-XXI/2023, handed down on 2 January 2025, which partially granted the judicial review petition against Law Number 7 of 2017 on General Elections, has so far only addressed restrictions on the manipulation of election participants’ ‘self-image’ in the form of photos or images. As a result, there is a significant legal vacuum regarding audio and video manipulation based on generative AI, such as deepfakes.
This legal uncertainty then gives rise to the liar’s dividend phenomenon. When technological fabrication becomes increasingly indistinguishable from reality, political actors who commit genuine violations can easily evade legal accountability by claiming that the evidence of their wrongdoing is merely AI fabrication or a hoax. This creates a systemic crisis of trust in the entire electoral process and its results.
TACTICAL STEPS NEEDED
Countering this asymmetric cyber threat requires multidimensional tactical steps. First, policymakers must promptly issue technical rules on digital campaigns requiring watermarking on all generative-AI content so the public can instantly identify authenticity. Second, electoral oversight authorities should be pushed to transform their conventional monitoring methods by integrating technological detection into the Election Vulnerability Index (IKP) as an early warning system.
Ultimately, however, the strongest defence lies in the hands of voters themselves. The cognitive resilience of civil society must be strengthened through grounded data literacy education. We must recognise that every digital activity of ours is a valuable commodity being fought over. A healthy democracy can only stand if the people’s right to vote is based on clear awareness, not on the output of algorithmic tuning.