On 1 October 2026, at an AI training session in Jakarta, Dwi Octavia Handayani, head of the DKI Jakarta women and child protection agency (PPAPP), warned about AI-made sexual deepfakes. According to Periskop.id's report of the event, she said Jakarta recorded 2,269 cases of violence against women and children in 2025. She also said Komnas Perempuan counted 1,091 technology-facilitated gender-based violence (KBGO) cases nationally, about 90% of them sexual. We could not independently match the 1,091 figure to Komnas Perempuan's published tallies, which other outlets report differently, so treat it as the official's claim, not a verified national count. The report does not discuss the legal framework; as background, Indonesia is often described as having no dedicated deepfake-pornography rules beyond Article 14 of the TPKS Law and the ITE Law.
The strongest case for new rules
The case for a dedicated deepfake law is serious. A victim of a synthetic sexual image suffers real reputational and psychological harm even though no intimate photo was ever taken. Prosecutors and judges may hesitate to stretch statutes written for recorded images or unwanted messages. Advocates also argue that a platform which lets users generate such images at scale should be held responsible, not just the individual who types the prompt. Those concerns are legitimate, and officials like Handayani are right to put numbers and victims in front of the public.
What Indonesian law already covers
The TPKS Law (Law No. 12 of 2022) is broader than the "no specific rules" framing suggests. Article 14 punishes electronic sexual violence: recording, photographing or screenshotting sexual content without consent, sending sexual electronic material to an unwilling recipient, and stalking someone electronically for sexual purposes. The base penalty is up to four years in prison and/or a fine of up to Rp200 million. Where the act is done to extort, threaten or coerce, it rises to six years and/or Rp300 million. Where the victim is a child or a person with a disability, the victim's consent is no defence.
Article 14 was written for recorded and transmitted content, so a fully synthetic image is a plausible gap in its text. That is a drafting question, and I am not aware of a published ruling that settles it. The state also has other tools. When Komdigi blocked Grok, it relied on Ministerial Regulation 5/2020 on private electronic system operators. Katadata reports that Article 9 of that regulation requires operators to ensure their systems do not host, facilitate or distribute prohibited content. It also reports that Komdigi cited the new Criminal Code (Law 1/2023, in force from 2 January 2026), whose pornography provisions carry penalties of six months to 10 years.
The Grok episode as a test case
Komdigi announced the temporary block of Grok on 10 January 2026. Minister Meutya Hafid said the government views non-consensual sexual deepfakes as "a serious violation of human rights." Indonesia was reported as the first country to deny access to the tool. Komdigi's reasoning, per Katadata, was that Grok lacked adequate safeguards against producing pornographic content from real photos of Indonesians, and that Grok was not registered as a private electronic system operator, though X was.
The block did not last. On 31 January, Komdigi normalised access conditionally after X Corp submitted a written commitment. X said it had added layered safeguards, restricted certain features, tightened internal rules and activated an incident response protocol. Director General Alexander Sabar said the commitment "is the basis for evaluation, not the end of oversight," and that the ministry would not hesitate to cut access again. The episode ran 21 days from block to restoration.
That sequence is a reasonable template. A temporary, reasoned, reversible intervention, tied to specific safeguard failures and lifted when the company committed to fix them, is proportionate. It acted against the tool producing the harm without a blanket ban on generative AI.
Where the real gap is
The evidence in this story points less to a missing statute than to a missing enforcement pipeline. Block orders address a platform in the abstract. Victims need takedown speed, a police unit that will take a complaint, and prosecutors willing to apply Article 14. Note also that Article 14 is a complaint-based offence unless the victim is a child or a person with a disability, so victims must file a report themselves, which is a real barrier in a stigma-heavy environment. Katadata notes victims can report to law enforcement or complain to Komdigi. Handayani's training-session framing, which focuses on digital literacy and women's awareness, addresses this side.
Our editorial view is that a new, deepfake-specific criminal law carries risks. Rushed definitions of "synthetic" or "manipulated" content tend to sweep in satire, political parody and journalism, and Indonesia's ITE Law has a record of being used against speech. A narrow amendment would be better: clarify that Article 14 reaches synthetic sexual imagery of an identifiable real person made without consent, and nothing more. Any platform duty should rest on notice-and-action timelines and documented safeguards, as the Grok conditions did, rather than on liability for all user output.
What to watch
- Whether any prosecution has applied Article 14 to an AI-generated image, which would settle the interpretation question without new legislation.
- Whether Komdigi publishes the test results it promised for X's safeguards, since it said it would verify them continuously.
- Whether the national KBGO figures cited at the event are reconciled with Komnas Perempuan's published data, so policy debates rest on one agreed number.
Deepfake abuse is real and growing, and Indonesia has shown it can act quickly. The next step is making existing law work for victims before adding new law that could reach speech it was never meant to.