10/03/2026
🇵🇬 PNG's AI Policy: A Good Start — But Is It Enough?
The Government recently released its draft AI framework for public services. I read it carefully. Here's my view.
First, credit where it's due.
This document is readable. It's written in plain language. It clearly tells citizens what Government will and won't use AI for. It says no secret citizen scoring. No AI-only decisions on your rights or benefits. No misuse of PNG cultural heritage.
Those are important promises and I'm glad they're in writing.
The architecture — building shared digital foundations first, then layering AI on top — is also the right approach. Think of it like building a house: you pour the foundation before you put on the roof. That logic is sound.
But i have concerns too.
This document is Part 1 of a two-part framework. Part 1 is the Citizen Guide — the public-facing document about trust and principles. Part 2 is the Government Implementation Pack — the operational detail, the standards, the rollout plans.
Part 2 is not yet public.
And that matters. Because without Part 2, we cannot answer the questions that actually determine whether this framework delivers anything real:
📌 Which services will be improved first, and by when?
📌 Which agency is accountable if something goes wrong?
📌 What are the actual targets — faster processing by how much? Fraud/Corruption detected at what rate?
📌 How will PNG build local AI capability so we're not permanently dependent on foreign systems?
📌 How exactly will PNG cultural heritage and our 800+ languages be protected — not just in principle, but in practice?
A policy without numbers is a wish. A commitment without a deadline is a suggestion.
The human review question — and why it cuts both ways.
The framework says AI will never make final decisions on your rights, benefits, or penalties alone. A human must always be involved. (HITL - Human In the Loop)
On the surface, that sounds protective. And for a first framework, it's the right starting position.
But here's the uncomfortable truth we also need to name: in many parts of our public service, the human in that loop is also the source of the delay, the inconsistency, and yes the corruption.
AI will only fix our system if we're honest about what's actually broken — and design around it, not just on top of it.
On sovereignty — and why it requires more than a policy statement.
The framework talks about protecting PNG's digital sovereignty — not allowing foreign AI systems to control our core government decisions. I support that principle completely.
But sovereignty requires engineers. It requires data scientists. It requires local technical capacity. A policy statement alone cannot make PNG sovereign in AI. We need a talent pipeline — UPNG partnerships, scholarships, diaspora recall, local developer ecosystems. That plan needs to be visible and funded.
My overall assessment:
Part 1 of this framework does its job. It builds public trust. It sets clear boundaries. It explains the vision in language ordinary Papua New Guineans can understand. That is not nothing — in fact, for a first public AI document, it's better than most.
But Part 1 by itself is not enough to prove PNG can deliver.
The real test is Part 2. When it is released, the questions we should all be asking are:
✅ Are there real targets with real deadlines?
✅ Are there named institutions with real accountability?
✅ Is there a pilot — one province, one service, proven before national rollout?
✅ Is there a genuine cultural heritage protection mechanism — not just a complaints pathway, but a consent and governance process built with community language custodians?
✅ Is there a talent and capability plan?
If Part 2 answers yes to those questions — this framework is genuinely credible and PNG should be proud of it.
If Part 2 is as general as Part 1 — then the hard questions need to be asked loudly, by citizens, by civil society, and by Parliament.
PNG deserves AI that works for us — not just AI that sounds good on paper.
I'll be watching Part 2 closely. I encourage everyone else to do the same.
