HSA consults on proposed exemption from manufacturer’s licensing and product registration requirements for AI-SaMD

25 June 2025

The Health Sciences Authority (“HSA”) is conducting a public consultation on a proposed exemption under the Health Products Act 2007. The consultation closed on 19 June 2025.?

Current regulatory framework?

Under regulation 3 and 6 of the Health Products (Medical Devices) Regulations 2010 (“Regulations”), specified healthcare service licensees that develop an Artificial Intelligence (“AI”) - Software as a Medical Device (“SaMD”) for in-house use (i.e. for their own patients) are not subject to the requirements for a manufacturer’s licence and product registration. This exception is designed to minimise the regulatory burden on licensees so that they can address the clinical needs of their patients more quickly. Healthcare service licensees operate within a regulated healthcare setting, subject to existing clinical governance frameworks, safety protocols, and oversight in the use of the SaMDs.?

The MOH Office for Healthcare Transformation Pte Ltd (“MOHT”), Synapxe Pte Ltd (“Synapxe”), and public healthcare clusters are increasingly leveraging technology to develop AI-SaMDs customised for Singapore’s public healthcare institutions to facilitate and improve the provision of patient care. While all part of public healthcare, they span multiple healthcare service licensees. Thus, HSA explains that the AI-SaMDs developed in public healthcare currently fall outside the scope of regulations 3 and 6 of the Regulations and cannot be easily deployed across public healthcare.?

Need for scaling of AI in public healthcare sector?

HSA states that AI-SaMDs that are developed by public healthcare for the purpose of public healthcare should be deployed not just in a single public healthcare institution, but across multiple institutions. This is so that public healthcare as a whole can reap the benefits of these customised AI-SaMDs, and also avoid duplicative costs if each institution has to develop its own AI-SaMD to perform the same function.?

Proposed exemption under Health Products Act 2007?

The objective of the proposed exemption under the Health Products Act 2007 is to run a regulatory sandbox for the deployment of AI-SaMDs developed by public healthcare for public healthcare, to assess the use of AI to provide public healthcare in an effective and safe manner.?

The proposed exemption would expand the current exception in regulations 3 and 6 of the Regulations to include AI-SaMDs that are:?

  • developed by MOHT, Synapxe, and public healthcare clusters and institutions;?
  • classified as class A (low risk) or class B (moderately low risk) under the Third Schedule to the Regulations;?
  • developed under the supervision and oversight of a clinician employed in a public healthcare institution holding the position of “Consultant” or higher;?
  • only for in-house use by public healthcare licensees; and?
  • notified to HSA at the point of deployment.

Non-compliance with these conditions would amount to breaches of the Health Products Act 2007, which are offences punishable by a fine of up to S$50,000, or imprisonment of up to two years, or both.?

Proposed additional safeguards and controls?

Under HSA’s proposal, MOHT, Synapxe, and public healthcare clusters and institutions developing these AI-SaMDs must self-attest annually to a quality management system (ISO 13485) and check against an internal product control checklist. The internal governance controls for public healthcare institutions, including cyber and data security requirements, would also continue to apply. The AI-SaMDs must be endorsed by the Chairman Medical Board or Chief Executive Officer of the public healthcare licensee before use.?

Post-market controls, including the requirement to report any adverse effect to HSA, would continue to apply, and breaches can be subject to HSA enforcement actions.?

HSA will monitor the proposed exemption over the course of 12 months to assess the regulatory sandbox. This 12-month duration may be extended to allow for a more robust assessment if there are insufficient indicators or?AI-SaMDs developed during the initial sandbox period.?

Reference materials?

The following materials are available on the HSA website :?