AnswersPublished22 sources
How Singapore Is Making Privacy-Preserving AI More Practical
Announced on 20 July 2026, Singapore’s PDPC and IMDA paired new privacy enhancing technology guidance with sandbox use cases. A healthcare project uses Trusted Execution Environments for cloud based AI on medical images, while Ant International’s case study combines several PETs for data collaboration.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Announced on 20 July 2026, Singapore’s PDPC and IMDA paired new privacy enhancing technology guidance with sandbox use cases.
- A healthcare project uses Trusted Execution Environments for cloud based AI on medical images, while Ant International’s case study combines several PETs for data collaboration.
- For businesses, the choice depends on the task: collaborate on model training while data stays local, or generate a dataset for sharing and analysis.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How are Singapore’s Personal Data Protection Commission (PDPC) and Infocomm Media Development Authority (IMDA) advancing privacy-preserving. Article summary: PDPC and IMDA are combining practical privacy-enhancing technology (PET) guidance with sandbox experimentation to make data-driven innovation easier without requiring unrestricted access to sensitive data; these initiati. Topic tags: general, general web, user generated, government, academic. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, water
Singapore’s Personal Data Protection Commission (PDPC) and Infocomm Media Development Authority (IMDA) announced new privacy-enhancing technology (PET) guidance and sandbox use cases at the Singapore Data Festival on 20 July 2026. The initiatives address a practical AI problem: organisations may have useful data but be unable or unwilling to pool sensitive records. 1
38
Two approaches to using data with greater privacy
PETs can support collaboration without sharing original data, or transform data to make sharing safer. The approaches solve different problems; choosing one starts with deciding whether the goal is joint model development or access to a dataset. 2
Federated learning: train across data that stays local
With federated learning, organisations train a model across decentralised datasets while keeping the underlying records in their local environments. Participants exchange model updates or aggregated insights rather than simply combining their raw datasets. 18
32
This can make collaboration possible when data is held by different organisations. It does not, by itself, establish that every model or information exchange is risk-free: the system still needs appropriate privacy, security and governance safeguards. 16
Synthetic data: create a dataset for sharing or analysis
Synthetic-data generation produces artificial data rather than moving the original records between organisations. It can serve a different need from federated learning: working with a dataset, rather than training a shared model while data remains distributed. PDPC’s synthetic-data guide focuses on structured data. 48
Synthetic data should not be assumed to be anonymous simply because it is artificial. Organisations still need to assess whether the generated data could reveal information about people and whether it is suitable for its intended use. More broadly, PDPC has cautioned that PETs do not replace sound governance, security, access controls or human oversight. 16
What the new guidance offers
The Singapore Data Festival announcements included a new Federated Learning guide and an updated Guide on Synthetic Data Generation. The guidance is intended to help organisations put PETs into practice; reporting on the federated-learning guidance describes adoption roadmaps and risk-management recommendations. 1
37
48
The practical choice is use-case driven: federated learning is relevant when participants want to develop a model together without pooling their original data; synthetic data is relevant when a shareable dataset is needed. Neither method is a universal substitute for checking privacy risks, data quality, security and operational fit.
Sandbox projects put PETs to the test
IMDA’s PET Sandbox, introduced in 2022, provides a setting for businesses to pilot privacy-enhancing technologies with support from the authority. 6
46 New use cases announced in 2026 show how different tools can be applied to healthcare and payments.
38
- Medical images: Singapore General Hospital is using Trusted Execution Environments for cloud-based AI inferencing on medical images, according to the announced use case. The project illustrates an effort to process sensitive material in a protected environment; its announcement should not be read as proof of particular clinical or cost benefits.
15
38
- Payment collaboration: Ant International’s case study describes combining federated learning, Multi-Party Computing and Homomorphic Encryption. The organisations involved can identify common customers while keeping underlying personal data in their own environments and co-training a prediction model.
These examples also underline why pilots matter: a PET needs to work for the specific collaboration, not just offer an appealing privacy label. Testing can help organisations assess whether an approach is suitable before wider deployment. 6
46
The takeaway for organisations
Singapore’s approach links guidance with practical experimentation: use federated learning when the goal is collaboration across data that remains local; consider synthetic data when a generated dataset is the more useful output; and test the technical and privacy assumptions in the intended setting. PETs can reduce the need to expose original data, but they are one part of responsible data use—not a blanket guarantee of privacy. 2
16
38
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "How Singapore Is Making Privacy-Preserving AI More Practical"?
Announced on 20 July 2026, Singapore’s PDPC and IMDA paired new privacy enhancing technology guidance with sandbox use cases.
What are the key points to validate first?
Announced on 20 July 2026, Singapore’s PDPC and IMDA paired new privacy enhancing technology guidance with sandbox use cases. A healthcare project uses Trusted Execution Environments for cloud based AI on medical images, while Ant International’s case study combines several PETs for data collaboration.
What should I do next in practice?
For businesses, the choice depends on the task: collaborate on model training while data stays local, or generate a dataset for sharing and analysis.