What Changed — The July 27 Digital Medical Product Notice Amendment
On July 27, 2026, Korea's Ministry of Food and Drug Safety (MFDS) amended its Notice on the Approval, Certification, Notification, Review and Evaluation of Digital Medical Products. This amendment does not create new rules for updating AI medical devices. It mostly aligns the Notice with higher-level regulations that had already changed. The sequence looks like this:
Updates That Need a Change Approval — and Those That Don't
The core of the negative approach: if a change is not significant, you manage it yourself without a change approval. According to press reports, AI retraining also needs no change approval unless it counts as a significant change. In practice, a first-pass assessment might look like this:
| Type of update | Likely outcome |
|---|---|
| UI improvements, bug fixes, refactoring with no performance impact | Likely self-managed |
| Retraining with added data within the same intended use and input data scope | Self-managed after checking it is not a significant change |
| Adding new input types, such as a new imaging device or modality | Likely a significant change |
| Expanding indications or changing intended use (e.g., diagnosis to prediction) | Significant change |
| Changing the algorithm structure (e.g., replacing the model architecture) | Likely a significant change |
These are examples only. The Enforcement Rule text is the final authority. Self-managed changes still have to be documented and validated. If you cannot produce that evidence during a post-market inspection, the burden only grows.
Change Management Plans — Defining Update Scope at Approval
For digital medical devices, a change management plan can be approved together with the device. It describes anticipated software changes, how they will be validated, and their impact on safety and effectiveness. Changes within the approved scope can then go into the product without a separate approval. The idea mirrors the US FDA's Predetermined Change Control Plan (PCCP).
A plan should cover at least:
Confirm which products are eligible and which forms to use in the current MFDS guidance.
Practical Preparation for Small Medical AI Firms
How KITIM Can Help
KITIM advises on change classification SOPs and change management plans, and provides regulatory strategy consulting. We also help digital health and medical AI companies connect with relevant government R&D programs. If you are unsure whether your next model update needs a change approval, contact the KITIM consulting team. We will work with you on decision criteria and a preparation timeline that fit your product.