AI Healthcare Compliance in Singapore: Key regulations to consider

Ministry of Health (MOH) in Singapore has invested around $200 million over the 5 years to support development and test-embedding of innovations in public healthcare institutions, including innovations in AI. Singapore took a use case approach, with more than ten such system-wide use case ranging from automating note-taking and clinical coding to predicting outpatient attendance.  

However, integrating AI in the healthcare system in Singapore is a tough nut to crack. A survey conducted by PMC (2025) reported that 87% of respondents feared that AI in healthcare would lead to increased data collection and mass surveillance. So how to build AI software applications responsibly to ensure data protection? This guide will break down some key regulations in Singapore, providing readers with best practices to build AI systems safely.  

Key takeaways 

• Misnistry of Health in Singapore has invested around $200 million over the 5 years to support innovations in public healthcare institutions, including AI innovations.  

• PMC (2025) found that 87% of respondents feared that AI in healthcare would lead to increased data collection and mass surveillance.  

• 70% of AI projects fail at data, not the model. Medical data is often scattered across multiple sources (EHR/EMR), making it hard for AI system to perform well.  

• In Singapore, AIHGle 2.0 provides practical guidance to support the safe development, deployment, and the use of AI in healthcare, benefiting patients and increasing trust. 

• Building AI responsibly means building AI systems that are safe, reliable, fair, transparent, and compliant with regulations in Singapore. 

• If the AI development vendor only makes general commitment that their solutions will be "compliant" without explaining the basis for it or applying the same compliance checklist to all applications, it's better to reconsider. 

Why AI healthcare compliance matters in Singapore  

Singapore takes healthcare seriously, treating it as a national priority and a key area for technological innovation. This country has set sights to be a global hub for developing and deploying AI solutions. It identifies five National AI projects focusing on key sectores, including healthcare. This creates a strong foundation for AI adoption across the healthcare industry. 

Healthcare possesses a huge amount of patient data, and if not managed properly and securely, it can lead to serious consequences. Patient data is shared across systems within a healthcare institutions and between public health organizations in healthcare. Also, when AI is integrated, it needs data to perform, which can increase the risk of data breaches. Notably, in 2024, around 1.5 million patients have had their non-medical personal particulars illegally accessed and copied, including NRIC number, address, gender, race and date of birth. 

Patients need to know how their data is collected, used, and stored. Following a clear framework or regulation can give patients a clear view of their data and increase trust among them.  

AI healthcare compliance challenges  

Many software development companies do not understand or have no experience building AI solutions for Singaporean clients, so they face multiple challenges when building AI that complies with complex regulations of this market. Here are several common challenges:  

Data engineering for AI  

70% of AI projects fail at data, not the model. Medical data is often scattered across multiple sources (EHR/EMR), with inconsistent formats, lacking a unified system and structure. Before incorporating AI imaging, the data needs to be collected, cleaned, and standardized. If the data is not properly input, the AI will not function effectively.  

Integration with legacy healthcare system  

Many healthcare facilities are still using outdated software systems that were not designed to be compatible with modern technologies like AI. Therefore, integrating AI into existing systems often requires extensive customization and technical adjustments to ensure that the systems can connect, exchange data, and operate synchronously. 

Data training for AI  

AI models rely on big data to learn and make predictions. However, healthcare data is often fragmented across systems, stored in inconsistent formats, or contains sensitive patient information. Preparing training data requires cleaning, standardisation, and validation to ensure the model performs reliably in clinical settings. 

Cross-border data transfer  

When outsourcing AI development to an offshore software development company, healthcare institutions may need to share patient data with external teams. However, transferring health data across borders raises concerns about data protection and regulatory compliance.  

To mitigate these risks, organisations need clear data-sharing agreements, strict access controls, and appropriate safeguards to ensure patient data remains secure throughout the development process. 

Key AI compliance in healthcare in Singapore  

1. PDPA in healthcare  

The Personal Data Protection Act (PDPA) establishes the baseline standard for protecting personal data in Singapore, governing how organizations collect, use, disclose, store, and protect personal data. For healthcare institutions, it means enabling patients to access the personal data held about them. Also, healthcare organizations must inform patients about how their data is used or disclosed over time.  

To have deeper insights into PDPA in healthcare in Singapore, take a look at our article where we explore how PDPA is applied to healthcare data security, privacy, and compliance. 

2. Artificial Intelligence in Healthcare Guidelines 2.0 (AIHGle 2.0)  

Singapore aims to adopt AI thoughtfully, prioritizing patient safety and clinical effectiveness. AIHGle 2.0 provides ethical guidelines for AI to support the safe development, deployment, and the use of AI in healthcare, which help to benefit patients and improve trust. This guideline emphasizes that AI should empower healthcare professionals in enhancing delivery, with patients at the heart of it.  

Key principles of AIHGLe 2.0:  

• Safety: Patient safety must be prioritized and must adhere to the fundamental medical principle of non-maleficence.  

• Fairness: AI in healthcare should be made accessible and should not result in discriminatory or unjust clinical outcomes.  

• Transparency: Every stakeholder should provide sufficient information for each stakeholder in the AI lifecycle.  

• Explainability: AI solutions should have appropriate levels of explainability tailored to user needs and associated risks. 

• Robustness: AI solutions should perform consistently under different deployment conditions. 

• Security and data protection: AI solutions should be secure-by-design. 

• AI alignment to human values or goals to shape AI objectives and adhust AI functions.  

What it means for AI solutions development:  

AI solutions developers, often AI development companies that develop, integrate, or maintain healthcare AI software, have to understand the regulatory requirements and ensure compliance for the AI solutions they build.  

Deployers, who deploy AI solutions to power healthcare delivery, should maintain operational readiness and make sure that healthcare staff are qualified and well-trained in AI use. Also, they need to implement governance framework focusing on risk management and quality assurance. Deployers are licensed under the Healthcare Services Act 2020 (HCSA).  

Users, who use AI solutions to support clinical work (often the healthcare organizations), are responsible for delivering safe and effective patient care. Also, they need to comply with relevant obligations and standards.  

3. AI Verify Foundation  

Developed by IMDA, AI Verify Foundation is an AI governance testing framework and software toolkit. It is built around 11 ethical principles and aligns with international AI governance frameworks. AI Verify supports organizations (including healthcare firms) in validating AI systems performance through standardized tests, both technical tests and process checks.  

How AI Verify Foundation supports healthcare firms: The AI Verify toolkit can be integrated into the healthcare organizations, allowing teams to run technical tests on their AI models and generate reports. It can perform technical tests on common supervised-learning classification and regression models for most tabular and image datasets.  

Moreover, it is an open-source toolkit, hosted on GitHub and governed by the AI Verify Foundation community.  

In addition to general regulations related to the responsible development of AI in Singapore, organizations should pay more attention to specific regulations for the healthcare sector.  

How to build AI responsibly in healthcare: Technical practices  

Building AI responsibly means building AI systems that are safe, reliable, fair, transparent, and compliant with regulations in Singapore. Here are several key ethical guidelines for AI.  

Data quality validation  

AI software applications in healthcare are only as reliable as the data they use. Before using data to train an AI model, healthcare organizations should make sure it is accurate, complete, and consistent. Missing information, duplicate records, or outdated medical data can affect how well the AI performs and lead to unreliable results. 

Data de-identification  

As patient data contains sensitive personal information, it is necessary to remove or mask identifying details before using it for AI development whenever possible. This helps protect patient privacy, although additional safeguards may still be needed to prevent re-identification. 

Encryption and access control  

Healthcare data needs to be protected both when it is stored and when it is transferred between systems. Use encryption and restrict access to authorized personnel only to reduce the risk of data leaks or unauthorized access. 

Reduce bias in training and clinical safety  

AI models can produce biased results if their training data does not represent different patient groups well. Use diverse datasets and test the model across relevant populations to identify potential bias and reduce the risk of unsafe clinical recommendations. 

Keep human-in-the-loop   

AI should support healthcare professionals, not replace their clinical judgment. Doctors and other qualified staff should be able to review AI-generated results, correct mistakes, and make the final decisions when patient safety is at stake. 

Monitor model performance  

An AI model may not perform as expected once it is used in real healthcare settings. Regularly track its accuracy, errors, and potential bias, and investigate any unexpected results. If the model needs updating or retraining, test it again before putting the changes into practice. 

It is notable that these practices are for reference only and do not cover all necessary requirements. Applying these practices DOES NOT mean that a business can guarantee its AI solutions are developed and deployed with absolute security. 

AI compliance checklist to assess a responsible AI vendor  

When choosing an AI development company to build AI solutions for the healthcare sector in Singapore, healthcare firms need to consider multiple factors to select the right one. Here are some common criteria that you can take a look at:  

Healthcare regulatory expertise  

As we mentioned before, AI development in healthcare has to follow strict regulations relating to patient data safety. That's why an AI development company needs to have a deep understanding of these requirements. You should choose vendors with many years of experience developing AI solutions for the healthcare sector by checking their website, portfolio, and customer feedback.  

If the software development company only makes a general commitment that their AI solutions will be "compliant" without explaining the basis for it or applying the same compliance checklist to all applications, it's better to reconsider.  

Data privacy and protection  

Choose a vendor with clear policies for handling patient data, including data minimization, de-identification, and retention. They should explain how patient information will be used, whether it may be used for AI training, and how data access and sharing are controlled.  

Data security  

A responsible AI vendor should protect healthcare data with encryption, access controls, and activity logs. They should also conduct regular security testing, address vulnerabilities, and maintain incident response procedures. If a vendor cannot explain its security measures or provide evidence of how patient data is protected, consider it a warning sign. 

AI risk assessment  

Choose a vendor that identifies potential AI risks early in development, including inaccurate outputs, hallucinations, and unexpected model behavior. They should assess the potential impact of these issues and establish safeguards to reduce harm. You should be cautious of vendors that focus only on model performance without considering the consequences of AI failures. 

Clinical validation and safety  

It is necessary to test AI using data and scenarios relevant to its intended clinical use. They should involve healthcare professionals in evaluating results, identify model limitations, and establish safety measures before deployment. Avoid vendors that rely solely on demos or general accuracy scores without providing relevant validation evidence. 

Bias and fairness testing  

Choose a vendor that evaluates whether its AI performs consistently across relevant patient groups. They should use representative datasets, measure differences in error rates, and address potential bias. Be cautious of vendors that report only overall accuracy without assessing performance across different populations or acknowledging limitations in their training data. 

Responsible AI track record  

Look for vendors with relevant project experience and evidence of responsible AI practices, such as testing reports, risk assessment procedures, or documented case studies. They should be transparent about previous challenges and how they addressed them. Be cautious of vendors that make broad claims about ethical AI without providing supporting evidence. 

The above are common criteria that healthcare facilities can consider before making a vendor selection decision. If hiring an offshore dedicated development team in offshore countries (such as Vietnam), healthcare firms can list a few potential vendors and then work more closely with them to clarify their security practices. 

Where Adamo APAC fits  

Adamo APAC provides high-quality AI software development solutions for companies across the Asia Pacific region. With more than 8 years of delivering software solutions, we have a deep understanding of the regulations in the targeted markets, including Singapore.  

With Adamo APAC, compliance is standard, not an upsell. Our AI solutions are built with compliance and responsible practices:  

• Holding ISO 27001 and ISO 9001:2015 certifications.  

• Every engagement includes PDPA-aligned data handling, robust NDAs signed within 24 hours and full IP and source code transferred to client from day one.  

• Our practices align with Singapore's IMDA Model AI Governance Framework and the new Model AI Governance Framework for Agentic AI. 

More importantly, Adamo APAC focuses on healthcare industry in Singapore. Our healthcare work spans Telehealth & Telemedicine, EHR/EMR Integration, Remote Patient Monitoring, Hospital & Clinic Management, mHealth, and Healthcare Data Analytics. We also integrate AI into healthcare workflow, enhancing clinical efficiency. Our AI adoption is built with validation and oversight, which supports the clinical judgment, not replacing it.  

Let's talk to us to build AI healthcare software that is compliant, secure, and reliable.