Ketika Negara Meminta Platform Mengetahui Umur Pengguna

Internet lama dibangun dengan satu asumsi nyaman:

platform tidak perlu benar-benar tahu siapa lo.

Bikin username.

Email.

Tanggal lahir.

Kalau ditanya usia, lo jawab sendiri.

Selesai.

Pada 2026, asumsi itu mulai berubah.

Indonesia melalui PP TUNAS mendorong platform lebih serius mengenali usia pengguna.

Wamenkomdigi Nezar Patria pada 30 Januari 2026 mengatakan platform perlu menerapkan sistem yang dapat mengenali usia pengguna sebagai bagian tanggung jawab perlindungan anak.

Beberapa hari kemudian, pemerintah juga menyorot fakta bahwa deklarasi tanggal lahir gampang dimanipulasi dan mendorong penggunaan pendekatan age inferential atau deteksi berbasis perilaku.

Pada Juli 2026, pemerintah kembali menyebut tantangan bahwa tiga dari lima anak dalam survei yang dirujuk memalsukan usia agar bisa mengakses media sosial.

Kepmen Komdigi Nomor 142 Tahun 2026 juga memasukkan mekanisme pengumpulan usia, memastikan usia, dan verifikasi usia sebagai bagian penilaian.

Jadi pertanyaannya bukan lagi: perlukah platform tahu umur?

Untuk kategori layanan tertentu, pemerintah memang mengarah ke kewajiban yang lebih kuat.

Pertanyaan yang lebih penting:

seberapa banyak platform perlu tahu?

Mengetahui umur tidak harus mengetahui identitas penuh

Ini distinction fundamental.

Platform bisa membutuhkan satu fakta: user di bawah 16 atau tidak.

Untuk itu, apakah perlu tahu NIK?

Nama legal?

Alamat?

Tanggal lahir exact?

Tidak selalu.

Age assurance should answer minimum question.

“16+ verified.”

No more.

Privacy-preserving credential bisa membantu.

Third party verifies.

Platform gets token.

Raw identity stays elsewhere or deleted.

This reduces risk.

Kalau setiap platform menyimpan KTP, problem baru.

Identity breach.

Age protection berubah surveillance.

Negara harus menetapkan principle minimization

Kalau hukum meminta platform tahu umur, hukum juga harus memastikan platform tidak overcollect.

Purpose: child safety.

Data yang dikumpulkan harus sesuai.

Tidak digunakan untuk ad targeting.

Tidak digunakan membuat family profile.

Retention terbatas.

Access controlled.

UU PDP sudah memberi principle data protection.

PP TUNAS harus berjalan selaras.

Child safety bukan exception untuk mengumpulkan semua data.

Justru data anak perlu proteksi tinggi.

Age declaration masih bisa punya tempat

Tidak semua layanan perlu verification berat.

Low-risk website.

News.

Education.

Age declaration mungkin cukup.

Risk-based assurance.

High-risk social feature needs stronger.

Payment.

Adult content.

Direct stranger interaction.

Higher assurance.

Regulator sebaiknya membuat ladder.

Level 1: self declaration.

Level 2: account signal.

Level 3: third-party verification.

Level 4: strong identity where justified.

Not one method all.

This keeps internet usable.

Age inferential menarik tapi berbahaya jika overused

Platform bisa melihat behavior.

Jam aktif.

Content watched.

Language.

Friend network.

From that infer: maybe child.

Useful for detecting fake age.

But this is profiling.

If done deeply, privacy issue.

Child protection can become excuse for behavioral surveillance.

Need limits.

What signals?

How long stored?

Can user challenge?

False inference?

If adult is classified child, feature blocked.

Need appeal.

If child classified adult, protection fails.

Accuracy matters.

Face age estimation also probabilistic

Selfie.

AI estimates 14.

User actually 17.

Or reverse.

Model error.

Skin.

Lighting.

Camera.

Population.

Need local validation.

Don’t make estimate sole truth for high-impact restriction.

Alternative method.

Manual review.

And raw biometric should not be retained unnecessarily.

Wajah bukan password.

If leaked, cannot change.

Age assurance vendors become powerful

Platform may outsource.

Specialized company verifies millions user.

This vendor gets: document, face, age, possibly guardian relation.

Critical data.

Need governance.

Security.

Audit.

No secondary use.

Data deletion.

Cross-border.

Incident.

If business model is free because data monetized, red flag.

Age assurance should be trust service.

Not ad-tech extension.

Universal verification can harm anonymity

Adult users sometimes need pseudonymity.

Political discussion.

Health.

Support groups.

Sexual violence survivors.

If every website demands identity to prove adult, chilling effect.

Age verification should not become real-name internet.

Prove attribute, not identity.

“Over 18.”

No legal name necessary.

This architecture is technically possible.

Policy should encourage.

Logged-out access creates challenge

No account.

How know age?

Platform might ask every session.

Annoying.

Or default child-safe.

Maybe best for some high-risk content.

User who wants adult feature verifies.

This is better than tracking device invisibly.

Default conservative.

Escalate when needed.

Again proportional.

Government itself should not gain unnecessary visibility

If national identity service provides age token, does government learn which website user visits?

Danger.

Architecture should avoid central authority seeing every verification transaction if possible.

Privacy by design.

Selective disclosure.

Decentralized credential.

No central browsing log.

Digital identity can help without becoming tracking.

Indonesia is also building IKD and broader digital public infrastructure.

If age assurance eventually connects, governance must be careful.

The same identifier should not follow user everywhere.

Children need transition as they age

User turns 16.

Restriction changes.

Turns 18.

Child status ends.

System should update.

Do not retain child label forever.

Age credential expiry.

Revalidation.

Remove guardian link when no longer needed.

Lifecycle.

Data minimization includes deletion over time.

Parental consent doesn’t last automatically forever.

Age progression matters.

Platforms should explain verification clearly

Why requested?

What data?

Who processes?

Stored how long?

What if fail?

Simple.

Not legalese.

If user sees selfie request without explanation, distrust.

Rightfully.

Transparency improves adoption.

For children, language age-appropriate.

For parents, clear.

Don’t hide behind “for safety.”

Explain.

Age verification error needs remedy

Platform says: “we think you are under 16.”

User is 20.

What now?

Appeal.

Alternative proof.

Fast.

False positive can lock adults.

False negative exposes children.

Both matter.

Regulator should monitor error rate.

Not just existence.

Age assurance quality.

Kepmen 142/2026 risk framework makes this type of operational assessment increasingly relevant.

Platform self-assessment should contain evidence.

No “we have AI age detection, done.”

Testing.

Accuracy.

Bias.

Spoof resistance.

Deepfake creates new challenge

AI can make fake face.

Photo older person.

Synthetic ID.

Age verifier attacked.

Need anti-spoof.

Liveness.

But stronger anti-spoof often collects more biometric.

Trade-off.

Security versus privacy.

No perfect point.

Risk-based.

High-risk action stronger.

Low-risk lighter.

Technology evolves.

Regulation should not specify one permanent method.

Set outcome.

Verifier must reasonably distinguish.

Allow innovation.

Platform must avoid age data for personalization beyond safety

This is crucial.

Once platform knows child is 14, advertiser would love use.

No.

Age data gathered for safety should not automatically feed marketing.

Purpose limitation.

Maybe coarse age band needed to make experience appropriate.

But profiling for commercial exploitation should have strict boundary.

Child safety data should not become revenue data.

Otherwise incentive distorted.

There is also fairness issue for users without documents

Some people lack easy digital identity.

Rural.

Foreign resident.

Special legal status.

If no document, don’t exclude from basic internet.

Alternative verification.

Assisted process.

Conservative mode.

Digital inclusion.

Indonesia diverse.

System designed for Jakarta KTP-holder only will fail nationally.

Standards should be interoperable

If 20 platforms each ask selfie, bad.

Reusable age credential better.

But one centralized monopoly risk.

Maybe multiple accredited verifier with common standard.

Competition.

Interoperability.

User chooses.

Token.

This reduces document sprawl.

Government can set trust framework.

Not necessarily operate every check.

Age assurance could become digital public infrastructure layer

Interesting future.

Like authentication.

But must have strict governance.

Minimal data.

No tracking.

Open standard.

Independent oversight.

If done well, users verify once and share age attribute selectively.

If done poorly, every click linked to national identity.

The difference is architecture.

Policy choices now matter.

Public debate is necessary

Because this is not technical detail.

It changes anonymity on internet.

Parents.

Children.

Civil society.

Privacy expert.

Platform.

Security.

Need consultation.

Government goal legitimate: protect children.

But mechanism can create unintended surveillance.

Policy should be tested publicly.

No black box age regime.

Ketika Negara Meminta Platform Mengetahui Umur Pengguna, the state is changing a foundational assumption of the internet.

Anonymous or pseudonymous access becomes harder in some contexts.

That can protect children.

It can also expose everyone.

The best design is not “know every user.”

It is “know only what is necessary for the risk.”

If platform needs to know 16+, give it 16+.

Not full identity.

If low-risk service doesn’t need age, don’t collect.

Ada satu pertanyaan yang juga penting: siapa yang boleh melakukan age inference terhadap user yang tidak pernah meminta fitur anak?

Kalau platform memantau seluruh perilaku untuk menebak umur semua orang, skala profiling menjadi besar. Regulasi perlu memastikan metode seperti ini tetap necessary dan proportional.

Age inference seharusnya punya threshold. Signal lemah jangan langsung mengubah status akun. Bisa memicu request verification tambahan.

Misalnya sistem menduga user di bawah 16 karena pola konten. Jangan otomatis menutup akun. Beri notice: “kami perlu memastikan usia.”

User bisa memilih metode verifikasi yang sesuai.

Ini lebih fair daripada keputusan diam-diam.

Transparansi penting karena inferential system bisa salah.

Ada orang dewasa yang suka konten game anak. Ada orang tua yang memakai akun untuk anak. Ada anak yang menggunakan perangkat keluarga.

Behavior bukan identity.

Karena itu inference sebaiknya digunakan sebagai trigger untuk assurance, bukan selalu sebagai bukti final.

Platform juga perlu memisahkan safety inference dari advertising inference. Kalau sistem sudah menyimpulkan user kemungkinan anak untuk safety, data itu jangan kemudian digunakan untuk membuat segment komersial.

No.

Safety signal stays safety.

Purpose limitation harus tegas.

Kalau tidak, insentif platform menjadi problem: semakin banyak yang diketahui tentang umur, semakin besar peluang monetisasi.

Regulator perlu memastikan child protection tidak berubah menjadi age-targeting infrastructure.

Dan kalau platform memilih tidak melakukan inference sama sekali, mereka tetap harus membuktikan bahwa metode assurance lain cukup untuk risk produk mereka.

Tidak ada one-size-fits-all.

Yang dinilai harus outcome: apakah anak secara reasonable terlindungi dan apakah privacy user tetap dijaga.

Protection should be precise.

The more data a safety system requires, the more safety obligations that system itself creates.