Malaysia's Proposed AI Law: What Is Proposed and What Is Not
Malaysia does not yet have a dedicated AI law. On 10 July 2026, the National AI Office (NAIO) released a public consultation paper on proposed AI governance legislation, with written feedback due by 31 July 2026. Legal-firm summaries report that tabling in Parliament is targeted for early 2027. None of this is in force, and the final text may differ.
The short version
- The proposal is principle-based and risk-based, with responsibilities assigned by role across the AI lifecycle.
- It proposes a Central AI Authority, safeguards, incident reporting and regulatory sandboxes.
- Separately, the Securities Commission consulted in July 2026 on board oversight and disclosure of AI use by listed companies. That is a different process.
Proposed, reported and undecided
| Item | Status | What to do now |
|---|---|---|
| Central AI Authority | Proposed in the consultation paper | Watch for the final institutional design |
| Risk-based approach, roles across the AI lifecycle | Proposed | Know whether you develop, provide or use each AI system |
| Incident reporting | Proposed | Set up detection and escalation for AI incidents now |
| Regulatory sandboxes | Proposed | Relevant to developers testing new systems |
| Tabling in Parliament | Reported as targeted for early 2027, not confirmed | Follow official announcements |
| Penalties and specific high-risk lists | Not settled in what we reviewed | Do not assume figures from commentary |
Based on public summaries of the NAIO consultation, including a Baker McKenzie summary. Check the National AI Office's own paper for the authoritative text.
No-regret steps for any organisation
- Build an AI inventory. Every AI system in use, built or bought, including tools staff adopt themselves: purpose, data, owner, vendor and who is affected.
- Assign responsibility. Name an owner for each system and record what your vendors are responsible for.
- Sort by risk. Separate uses that can harm people, such as hiring or credit decisions, from low-risk productivity uses.
- Prepare an incident route. Decide what counts as an AI incident and who handles it.
- Align with what already applies. The PDPA and sector rules already cover much AI use that involves personal data.
Frequently asked questions
Is there an AI law in Malaysia now?
No dedicated AI law is in force. The National AI Office issued a public consultation paper on proposed AI governance legislation on 10 July 2026. Existing laws such as the PDPA and sector rules already apply to AI use that involves personal data, so organisations are not operating in a legal vacuum.
When will it take effect?
Nobody can say yet. Legal-firm summaries report that tabling in Parliament is targeted for early 2027, but that is a target, not a commitment, and a tabled proposal still has to pass and be brought into force. Plan for a direction of travel, not a date.
Who would it apply to?
The consultation proposes a risk-based approach with responsibilities assigned by role across the AI lifecycle, which suggests developers, providers and users could each carry duties. Exact definitions and exemptions are not settled, so treat any firm statement about who is covered as premature.
How is this different from the Securities Commission's consultation?
The Securities Commission consulted separately in July 2026 on board oversight and disclosure of AI use by listed companies. That concerns corporate governance for public listed companies. The NAIO proposal is a cross-sector AI governance framework. A listed company would need to watch both, because they address different things.