1. Introduction

Artificial Intelligence (AI) is increasingly being used in mental healthcare in Singapore to provide psychological assessments, mental health screening, emotional support, automated counselling, symptom analysis and therapy recommendations.

AI-powered mental health systems may help identify symptoms associated with depression, anxiety, bipolar disorder, post-traumatic stress disorder (PTSD), schizophrenia and other psychological conditions. However, incorrect diagnosis or inappropriate therapeutic recommendations can expose patients to serious physical, psychological and financial harm.

AI Mental Health Therapy Wrong Diagnosis Harm Claims concern civil liability arising when an AI-powered mental health platform, healthcare provider, psychiatrist, psychologist, hospital or technology developer causes injury through an inaccurate diagnosis, negligent treatment recommendation, failure to identify a serious mental health condition or inappropriate reliance on automated clinical decisions.

Singapore does not presently have a separate, comprehensive civil liability regime exclusively governing AI mental health misdiagnosis. Such claims would principally be examined through established principles of medical negligence, professional responsibility, contract law, product-related liability, personal data protection and psychiatric injury.

An important distinction is that an incorrect diagnosis does not automatically establish negligence. The claimant must establish the relevant legal duty, breach, causation and legally recognised damage.

2. Legal framework governing AI mental health therapy in Singapore

The principal legal frameworks include the following.

A. Common law of negligence

The tort of negligence is the principal route for claiming compensation when an AI-assisted mental health diagnosis causes injury.

A claimant generally needs to establish:

The defendant owed the claimant a duty of care.

The defendant breached the applicable standard of care.

The breach caused the claimant's injury.

The injury was not too remote.

The claimant suffered provable damage.

Potential defendants include:

Psychiatrists and other medical practitioners.

Hospitals and mental healthcare institutions.

AI mental health service providers.

Healthcare technology companies.

Organisations responsible for deploying and supervising AI systems.

B. Bolam–Bolitho standard of medical negligence

Singapore applies the Bolam test, subject to the Bolitho requirement of logical defensibility, in assessing professional standards for medical diagnosis and treatment.

Under this framework:

A professional's conduct is assessed against a responsible body of professional opinion.

The supporting professional opinion must be logically defensible.

Courts examine whether the professional exercised reasonable clinical judgment.

For AI-assisted mental healthcare, the relevant question would be whether the healthcare professional's reliance on the AI system was professionally reasonable in the circumstances.

A psychiatrist cannot necessarily escape liability by arguing that the AI system generated the incorrect diagnosis.

C. Personal Data Protection Act 2012

The Personal Data Protection Act (PDPA) is relevant where AI mental health systems collect, process, disclose or use personal data.

Mental health information can be particularly sensitive in practice.

Potential problems include:

Unauthorised disclosure of psychiatric records.

Insecure storage of counselling conversations.

Improper use of patient information to train AI models.

Inaccurate personal information.

Unauthorised sharing of psychological assessments.

A PDPA violation does not automatically establish a private damages claim for every resulting harm. The applicable statutory enforcement mechanisms and any independently established civil causes of action must be considered.

D. Health Products Act 2007

Where an AI system qualifies as a medical device or software medical device, the Health Products Act and applicable Health Sciences Authority requirements may become relevant.

The classification depends on the software's intended purpose and functionality.

A general emotional-support chatbot should not automatically be treated as a regulated medical device. An AI system intended to diagnose or guide the treatment of mental disorders may raise different regulatory questions.

E. Healthcare Services Act 2020

The Healthcare Services Act provides a regulatory framework for licensed healthcare services in Singapore.

Where AI is incorporated into a licensed healthcare service, the provider must consider its applicable licensing, clinical governance and patient-safety obligations.

The use of AI does not remove the responsibility of healthcare providers to maintain appropriate professional standards.

3. Common situations giving rise to AI mental health wrong diagnosis claims

KAIST Deploys AI to Streamline Psychiatric Intake | Let's Data Science

1. Incorrect psychiatric diagnosis

AI incorrectly identifies depression as ordinary stress or misclassifies bipolar disorder as unipolar depression, resulting in inappropriate treatment.

Same-Day Telepsychiatry Platforms November 2025 - Legion Health | Texas psychiatry covered by insurance - get treatment for anxiety today

2. Failure to identify serious mental health conditions

An AI therapy platform fails to recognise warning signs of suicidal behaviour, psychosis or severe depression.

Tess - ALLOne Health Mental Health Chatbot

3. Incorrect automated therapy recommendations

The system recommends inappropriate psychological interventions or treatment plans based on inaccurate assessments.

Clinical AI that is more honest about what it doesn’t know | WashU McKelvey School of Engineering

4. Excessive reliance on AI

A psychiatrist or psychologist accepts an automated diagnosis without conducting an adequate independent assessment.

Suicide Risk Factors, Warning Signs, and How to Help

5. Psychological injury caused by misdiagnosis

A patient experiences worsening psychiatric symptoms, emotional distress or deterioration following an inaccurate diagnosis or inappropriate therapy.

4. Six or more important Singapore case laws

Seven relevant judicial authorities

There is an important qualification: Singapore has not established a reported body of cases specifically concerning AI-powered mental health therapy misdiagnosis. The following decisions provide the established legal principles that would be relevant by analogy. They are not cases in which courts necessarily decided AI-related liability.

Case 1. Dr Khoo James and Another v Gunapathy d/o Muniandy and Another Appeal [2002] SGCA 25

Court: Singapore Court of Appeal.

Facts:

The patient alleged negligence against doctors who diagnosed her with a suspected brain tumour and recommended radiosurgery. The treatment resulted in serious complications, including radionecrosis and disability.

Legal issues:

Whether the doctors had negligently diagnosed the patient's condition.

Whether the recommended treatment was professionally appropriate.

Whether the medical opinions supporting the doctors' decisions were acceptable.

Judgment and legal principles:

The Court of Appeal applied the Bolam test, subject to the Bolitho requirement that professional medical opinion must be logically defensible.

The court recognised that medical professionals are not automatically negligent simply because another group of professionals might have adopted a different approach.

Relevance to AI mental health therapy:

Suppose an AI mental health platform identifies a patient as suffering from ordinary depression when the patient actually exhibits symptoms of bipolar disorder.

A psychiatrist who relies on the AI assessment may face liability if that reliance falls below the applicable professional standard.

The healthcare provider cannot simply rely on the fact that the software was developed by a reputable technology company.

Case 2. Hii Chii Kok v Ooi Peng Jin London Lucien and Another [2017] SGCA 38

Court: Singapore Court of Appeal.

Facts:

The patient underwent major pancreatic surgery following a medical diagnosis that he alleged was incorrect. He argued that the surgery was unnecessary and that the medical professionals had failed to provide appropriate advice.

Legal issues:

Negligent medical diagnosis.

Standard of professional care.

Medical advice and patient autonomy.

Disclosure of treatment alternatives and material risks.

Judgment and legal principles:

The Court of Appeal maintained the Bolam–Bolitho framework for diagnosis and treatment.

However, it established a more patient-centric approach to medical advice.

In particular, the court's approach to medical advice considers:

Whether the doctor was aware of the information relevant to the patient's decision.

Whether the doctor provided information that a reasonable patient in that position would regard as material.

Whether the doctor should have recognised that the particular patient would regard additional information as material.

Relevance to AI mental health therapy:

An AI-assisted mental healthcare provider should not mislead patients into believing that an automated diagnosis is necessarily accurate.

Patients should receive appropriate information about:

The limitations of automated assessments.

The availability of human psychiatric evaluation.

The risks of recommended therapy.

Relevant alternative treatment options.

The limitations of AI-generated clinical recommendations.

This case is particularly important when examining patient autonomy and informed decision-making in AI-assisted therapy.

Case 3. Noor Azlin bte Abdul Rahman v Changi General Hospital Pte Ltd and Others [2019] SGCA 13

Court: Singapore Court of Appeal.

Facts:

The patient repeatedly attended Changi General Hospital, where medical investigations identified an abnormality in her lungs.

The hospital failed to establish an adequate system for proper follow-up of relevant radiological findings. The delay eventually contributed to the progression of her lung cancer.

Legal issues:

Negligent diagnosis.

Hospital systems and patient management.

Failure to follow up on medical findings.

Causation arising from delayed diagnosis.

Judgment:

The Court of Appeal found that the hospital had breached its duty of care by failing to ensure appropriate follow-up.

The court also identified serious inadequacies in the hospital's patient-management arrangements.

Relevance to AI mental health therapy:

This case is particularly significant for AI-based mental health platforms.

Consider a system that repeatedly identifies a patient as experiencing severe depression but fails to trigger an appropriate clinical review.

If the healthcare provider has no proper monitoring, escalation or follow-up procedures, the provider may face negligence liability.

The case demonstrates that liability can arise from defective healthcare systems, not merely an individual professional's diagnostic mistake.

Case 4. Noor Azlin bte Abdul Rahman and Another v Changi General Hospital Pte Ltd and Others [2021] SGHC 10

Court: Singapore High Court.

Facts:

Following the Court of Appeal's decision in the earlier proceedings, the High Court considered the assessment of damages arising from the delayed diagnosis.

The patient's cancer had progressed, and she subsequently died.

Legal issues:

Assessment of damages.

Consequences of delayed diagnosis.

Medical causation.

Compensation following deterioration in health.

Judgment:

The High Court awarded damages of approximately S$326,620.61.

The decision illustrates how a claimant's losses may be assessed when negligent healthcare management causes a deterioration in the patient's medical condition.

Relevance to AI mental health therapy:

If an AI mental health misdiagnosis causes a patient to receive inappropriate treatment for an extended period, the patient may seek compensation for proven consequences, including:

Additional treatment expenses.

Costs of psychiatric rehabilitation.

Loss of earnings.

Pain and suffering.

Other legally recoverable losses.

The claimant must establish that the negligent conduct caused the claimed deterioration.

Case 5. Ngiam Kong Seng and Another v Lim Chiew Hock [2008] SGCA 23

Court: Singapore Court of Appeal.

Facts:

The case concerned negligence claims involving psychiatric harm arising from a road accident.

Legal issues:

Whether psychiatric injury can give rise to damages.

Whether the psychiatric injury was reasonably foreseeable.

Whether a sufficient legal relationship existed between the parties.

Judgment and legal principles:

The Court of Appeal explained that claims for psychiatric harm require consideration of:

A recognisable psychiatric illness.

Factual foreseeability.

Sufficient legal proximity.

Whether policy considerations negate the duty of care.

The Spandeck framework is applied when determining the existence of a duty of care.

Relevance to AI mental health therapy:

A patient who develops a recognised psychiatric illness following negligent AI-assisted therapy may potentially pursue a damages claim.

However, ordinary disappointment, temporary anxiety or dissatisfaction with an AI diagnosis will not automatically qualify as a compensable psychiatric injury.

Medical evidence is particularly important in establishing the nature and seriousness of the injury.

Case 6. Spandeck Engineering (S) Pte Ltd v Defence Science & Technology Agency [2007] 4 SLR(R) 100

Court: Singapore Court of Appeal.

Legal issues:

The case established Singapore's general framework for determining whether a duty of care exists in negligence.

Judgment and legal principles:

The court established a structured approach involving:

Factual foreseeability of harm.

Sufficient legal proximity between the parties.

Consideration of policy factors that might negate a prima facie duty.

Relevance to AI mental health therapy:

This framework is relevant when determining whether an AI developer, healthcare provider or platform operator owes a duty of care to a patient.

For example, a company providing a mental health diagnostic platform may have a sufficiently close relationship with users depending on its role, representations, contractual arrangements and involvement in the provision of healthcare.

However, a duty cannot be assumed merely because a company develops general-purpose AI software.

Case 7. Tiong Sze Yin Serene v Chan Herng Nieng [2022] SGHC 170

Court: Singapore High Court.

Legal issues:

The decision examined the principles governing medical negligence, including the duty owed by a medical practitioner, the standard of care and the Bolam–Bolitho approach.

Judgment and legal principles:

The court explained the elements of negligence and applied the established principles governing professional medical responsibility.

It also discussed the application of the Bolam–Bolitho test to diagnosis and treatment.

Relevance to AI mental health therapy:

The decision is useful when considering the responsibilities of medical professionals who incorporate automated systems into clinical decision-making.

A professional must still exercise appropriate judgment rather than treating an AI-generated result as an unquestionable clinical conclusion.

Case references: [2002] SGCA 25; [2017] SGCA 38; [2019] SGCA 13; [2021] SGHC 10; [2008] SGCA 23; [2007] 4 SLR(R) 100; [2022] SGHC 170.

5. Who can be held legally responsible for AI mental health misdiagnosis?

Liability depends on the role played by each participant and the applicable legal duty.

A. Liability of psychiatrists

A psychiatrist may face a negligence claim if they:

Rely blindly on an AI-generated diagnosis.

Fail to conduct an appropriate clinical assessment.

Ignore significant symptoms.

Fail to arrange necessary investigations.

Recommend unsuitable treatment.

Fail to monitor the patient's progress.

The fact that AI assisted the diagnosis does not eliminate the psychiatrist's professional responsibilities.

B. Liability of AI developers

An AI developer may potentially face liability where the claimant establishes an applicable duty and breach, for example through:

Defective diagnostic design.

Inadequate testing.

Failure to address foreseeable model limitations.

Misleading claims about diagnostic accuracy.

Failure to implement appropriate safety measures.

However, there is no automatic rule that an AI developer is liable whenever its software produces an incorrect psychiatric assessment.

C. Liability of hospitals and healthcare institutions

Healthcare institutions may face claims involving:

Poor AI procurement procedures.

Inadequate clinical governance.

Failure to train staff.

Inappropriate automation of psychiatric assessments.

Failure to establish human-review mechanisms.

Failure to respond to patient safety warnings.

The Noor Azlin litigation provides particularly useful guidance on institutional responsibility for inadequate patient-management systems.

6. Causation and proof of harm

Causation is one of the most important elements in AI mental health misdiagnosis litigation.

A patient must generally demonstrate that the defendant's negligent conduct caused the injury complained of.

For example:

Illustrative situation

A patient approaches an AI mental health platform because of severe depressive symptoms. The system incorrectly classifies the symptoms as mild anxiety. The patient receives unsuitable recommendations and does not obtain an appropriate psychiatric assessment. The patient's condition subsequently deteriorates.

The claimant would need to establish:

The AI assessment or its clinical use was negligent.

The healthcare provider or other defendant owed an applicable duty of care.

Appropriate professional assessment would probably have identified the relevant condition.

The delay or inappropriate treatment caused the deterioration.

The deterioration resulted in legally recognisable and provable damage.

Important categories of damage

A claimant cannot simply attribute every subsequent mental health problem to an AI misdiagnosis. Expert psychiatric evidence and a reliable causal connection are essential.

7. Contractual liability for AI mental health therapy

Apart from negligence, contractual liability may arise when an AI mental health platform provides services under an agreement with a patient.

Potential contractual problems include:

Failure to provide the promised services.

Misrepresentation of diagnostic capabilities.

Failure to provide agreed professional supervision.

Failure to comply with contractual safety commitments.

Inadequate delivery of promised clinical assessments.

The applicable terms and conditions, representations made to users and nature of the service will be relevant.

The Unfair Contract Terms Act 1977 may also be relevant when contractual provisions attempt to exclude or restrict liability for negligence, subject to its statutory requirements.

A platform cannot necessarily avoid responsibility through a disclaimer stating that all AI-generated information is provided at the user's own risk.

8. Product-related liability and defective AI systems

AI mental health software may raise product safety and product-related liability issues.

Potential defects include:

Design defects: The diagnostic model is inadequately designed for its intended clinical purpose.

Training-data defects: The model was trained using insufficiently representative data.

Testing defects: The system was not adequately validated before deployment.

Warning defects: Users are not properly informed about diagnostic limitations.

Monitoring defects: Developers fail to respond appropriately to known safety problems.

The Consumer Protection (Fair Trading) Act 2003 may be relevant where misleading or unfair practices are involved.

The Sale of Goods Act 1979 and applicable contractual or common-law principles may also become relevant depending on the transaction and the nature of the software.

It is important not to assume that every software malfunction automatically gives rise to a statutory product-liability claim. The legal classification of the software and the applicable cause of action must be established.

9. Evidence required to establish an AI mental health wrong diagnosis claim

A claimant should preserve evidence showing what the AI system did, what the healthcare professional knew and how the patient's condition developed.

Digital evidence may be particularly important because AI systems can change their outputs following software updates or model changes.

A claimant should therefore seek to preserve relevant records, including the version of the software used at the time of the incident.

10. Possible legal remedies in Singapore

A successful claimant may seek the following remedies, depending on the cause of action and available evidence.

A. Compensatory damages

Compensation may cover:

Medical expenses.

Psychiatric treatment.

Counselling and rehabilitation.

Loss of income.

Pain and suffering.

Future care requirements.

B. Special damages

These are financial losses that can be specifically identified and quantified, such as hospital bills, medication expenses and lost earnings.

C. General damages

These may compensate for non-financial injuries, including pain, suffering and loss of amenity.

D. Injunctive relief

In appropriate proceedings, a claimant may seek an injunction to restrain continuing unlawful conduct, subject to the relevant legal requirements.

E. Regulatory complaints

Depending on the conduct involved, complaints may be made to:

Singapore Medical Council.

Ministry of Health.

Health Sciences Authority.

Personal Data Protection Commission.

Regulatory complaints and civil damages claims are separate processes. A regulatory finding does not automatically establish every element of a private negligence action.

11. Defences available to AI healthcare providers

Defendants may raise several legal defences.

1. Reasonable professional practice

A healthcare professional may argue that the diagnostic decision was supported by a responsible and logically defensible body of professional opinion.

2. No breach of duty

The defendant may argue that the diagnosis was a reasonable clinical judgment based on the information available at the relevant time.

3. Lack of causation

The defendant may argue that the patient's deterioration would have occurred even if the diagnosis had been correct.

4. Pre-existing psychiatric conditions

The defendant may argue that some or all of the claimed injury arose from an existing mental health condition rather than the alleged negligence.

5. Contributory negligence

Where legally applicable, a defendant may argue that the claimant's own unreasonable conduct contributed to the damage. This is fact-sensitive and should not be assumed merely because a patient followed an AI recommendation.

6. Contractual limitations

A defendant may rely on applicable contractual terms, although exclusion clauses are subject to statutory controls and other legal limitations.

12. Regulatory and ethical governance of AI mental healthcare

Singapore's AI governance environment is relevant to determining whether reasonable safeguards were implemented.

Important governance considerations include:

Human oversight of automated clinical decisions.

Transparency concerning AI capabilities.

Accuracy and reliability testing.

Protection of patient confidentiality.

Prevention of algorithmic discrimination.

Monitoring of adverse outcomes.

Accountability for clinical decisions.

The Model AI Governance Framework and AI Verify provide useful governance references, although they do not independently establish a civil cause of action for compensation.

The Ministry of Health has also emphasised human oversight in AI-assisted clinical workflows. Its 4 August 2026 parliamentary response concerning AI-assisted clinical note-taking explained that healthcare professionals must review AI-generated notes before incorporating them into clinical records. This concerns note-taking rather than AI psychiatric diagnosis, but illustrates the importance of human accountability in healthcare AI.

13. Practical hypothetical cases

The following examples illustrate how Singapore's legal principles could apply.

14. Important legal principles emerging from the case law

The seven authorities discussed above establish several important principles for future AI mental health litigation.

1. Professional accountability: Healthcare professionals remain responsible for exercising appropriate clinical judgment.

2. Logical defensibility: Professional decisions must satisfy the Bolam–Bolitho standard where it applies.

3. Patient autonomy: Patients must receive appropriate material information to make informed healthcare decisions.

4. Institutional responsibility: Healthcare providers must maintain adequate systems for patient monitoring and follow-up.

5. Causation: An incorrect diagnosis alone is insufficient; the claimant must prove legally relevant harm caused by the breach.

6. Psychiatric injury: A recognised psychiatric illness and the applicable duty-of-care requirements are important when claiming psychiatric harm.

7. Compensation: Damages depend on the nature, seriousness and evidential proof of the injury.

15. Conclusion

Civil liability for AI mental health therapy wrong diagnosis harm claims in Singapore is likely to develop through the application of established medical negligence and tort principles to new technological circumstances.

The most important judicial authorities are Gunapathy, Hii Chii Kok, Noor Azlin, Ngiam Kong Seng and Spandeck Engineering.

Together, these decisions provide a foundation for determining professional responsibility, institutional negligence, patient autonomy, psychiatric injury and causation.

The central legal principle is that AI-assisted healthcare does not eliminate human and institutional accountability. Where a healthcare provider negligently relies on an inaccurate AI diagnosis, fails to supervise automated assessments or does not arrange appropriate clinical intervention, civil liability may arise if the claimant proves the necessary elements of the relevant cause of action.

Singapore's future legal development will likely require greater attention to AI diagnostic transparency, clinical validation, human supervision, software accountability and effective compensation for patients harmed by automated mental healthcare systems.

16. Key case law reference list

For academic research and further legal study:

Research qualification: These are relevant Singapore authorities on medical negligence, psychiatric injury and duty of care, not seven precedents specifically deciding AI mental health misdiagnosis. That distinction is important when using this material in academic writing or legal submissions.