Skip to content

Sri Lanka’s PDPA and AI: what to prepare before 2027

A practical checklist for Sri Lankan organisations using AI on personal data — data mapping, purpose, minimisation, residency, human review and audit logs — ahead of the PDPA’s core obligations.

By NextAura

3 min read

In short

Sri Lanka’s Personal Data Protection Act, No. 9 of 2022, has core obligations scheduled to take effect on 1 January 2027. Organisations using AI on personal data should prepare now: map what data each AI system touches, limit it to a clear purpose, decide where it is stored and processed, keep a human in the loop for important decisions, and log what the system does.

Key takeaways

  • Start with a data map: which AI systems use which personal data, and why.
  • Design for minimisation — send models only the fields a task needs.
  • Decide data residency early: your servers, a Sri Lankan data centre or an approved cloud region.
  • Keep people in charge of decisions that matter, and keep audit logs.

Why AI projects need PDPA thinking from day one

AI systems are hungry for data. A customer-service assistant reads messages, a document pipeline reads forms and IDs, and a knowledge assistant indexes HR or customer files. Each of these processes personal data, and each is far easier to make compliant when privacy is part of the design rather than a patch at the end.

The Personal Data Protection Act, No. 9 of 2022, is Sri Lanka’s first comprehensive data protection law. Its core obligations are scheduled to take effect on 1 January 2027, which gives organisations a short window to bring existing systems — and any new AI projects — into line.

A practical checklist for AI systems

  1. 1Map the data. For each AI use case, list the personal data it reads, stores and produces, where that data comes from and who can see it.
  2. 2Name the purpose. Write down what each system is for, and do not reuse the data for unrelated purposes without a lawful basis.
  3. 3Minimise. Mask or remove fields the model does not need — national ID numbers, phone numbers and account details rarely need to reach a language model.
  4. 4Choose where data lives. Decide whether processing happens on your own servers, in a Sri Lankan data centre or in a cloud region you approve, and check cross-border transfer requirements with your advisers.
  5. 5Set retention. Decide how long prompts, transcripts, documents and logs are kept, and delete them on schedule.
  6. 6Keep a human in the loop. Route decisions with real impact on people — credit, claims, hiring — to a person for review.
  7. 7Log and monitor. Record what the system did and why, so you can answer questions from customers, auditors or the regulator.
  8. 8Support people’s rights. Make sure you can find, correct or delete an individual’s data across AI systems, not only in your core database.

What this looks like in a real build

In practice, PDPA-aware AI is mostly good engineering: role-based access so an assistant only retrieves documents the user may see, redaction before data leaves your network, configurable retention, and dashboards that show what the system has done. We cover our approach on the Security & Trust page.

This article is general information, not legal advice. Requirements and timelines can change — confirm how the PDPA applies to your organisation with your legal adviser and the Data Protection Authority of Sri Lanka.

Where to start

If you already run chatbots, automation or analytics on customer data, start with an inventory and a risk check of those systems. If you are planning a new AI project, build the checklist above into the discovery phase. Our AI Readiness Check takes about three minutes and shows where your data and governance foundations stand.

Frequently asked questions

01When does Sri Lanka’s PDPA take effect?

The Personal Data Protection Act, No. 9 of 2022, has core obligations scheduled to take effect on 1 January 2027. Check current timelines with your legal adviser, as dates can be amended.

02Does the PDPA apply to AI chatbots?

If a chatbot collects or processes personal data — names, phone numbers, account details or messages about individuals — the processing is covered like any other. Design it with a clear purpose, minimal data, retention limits and access control.

03Can AI data stay in Sri Lanka?

Yes. AI systems can be deployed on your own servers, in a Sri Lankan data centre, or in a cloud region you approve, depending on the models and services used.

Let’s put this to work.

Tell us about the process you want to improve or the product you want to build. Every message is read and answered personally.

hello@nextaura.lk · Colombo, Sri Lanka