When Machines Malpractice: Liability For AI And Robotic Healthcare In Australia

The rapid integration of artificial intelligence (AI) and robotics into healthcare promises significant advancements in diagnostic accuracy and therapeutic effectiveness, however, it simultaneously poses substantial challenges to established medical malpractice frameworks.

This article was originally published in the Lexis Nexis Health Law Bulletin Volume 34 No 3.

Introduction

In recent years, Australia has seen a rapid uptake of AI across many areas of healthcare. In response to the growing integration of AI in clinical settings, the Australian Government released its 2025 Final Report on Safe and Responsible AI in Health Care,¹ acknowledging that existing legislation is not fully equipped to manage the risks posed by AI in high-stakes clinical environments.² The report calls for national policy leadership, tailored regulatory frameworks and mandatory guardrails for high-risk AI applications.³ It also emphasises the need for clear accountability structures, particularly in scenarios where AI systems influence or automate clinical decisions.⁴

As Australia begins implementing national recommendations, the implications of AI integration in healthcare are becoming increasingly urgent, particularly for clinical, legal and insurance professionals navigating care delivery, risk management and compliance. It introduces new challenges, especially for liability in cases where, for example, an AI system misguides a procedure, as well as informed consent when AI is involved in clinical decision-making.

Legal liability

As Australia begins implementing national recommendations, the implications of AI integration in healthcare are becoming increasingly urgent, particularly for clinical, legal, and insurance professionals navigating care delivery, risk management, and compliance. It introduces new challenges, especially for liability in cases where, for example, an AI system misguides a procedure, as well as informed consent when AI is involved in clinical decision-making.

AI introduces new complexities in determining the responsible party when negligence occurs. Traditionally in healthcare, hospitals and individual clinicians owe a duty of care to their patients, pursuant to common law and civil liability legislation. This duty persists regardless of whether AI decision-support tools have been utilised. The growing use of AI is prompting a redefinition of what constitutes the 'standard of care'. As with other medical equipment, clinicians must ensure responsible use of AI as part of patient care. Although software developers and IT system integrators might theoretically owe duties directly to patients if their products cause foreseeable harm, federal laws typically channel such claims through statutory product liability under the Australian Consumer Law (ACL),⁵ rather than through negligence.

As a result, plaintiffs will generally turn first to the healthcare providers using AI-integrated systems as accountable parties in negligence, with courts unlikely to accept sole reliance on AI recommendations as a defence against clinical responsibility.⁶

However, highly autonomous AI scenarios, where AI acts independently, can complicate liability by requiring the identification of an accountable human or institution behind the AI system. In these scenarios, responsibility will likely rest with either the healthcare institution deploying the AI or individual clinicians supervising its use. In circumstances of vicarious liability, or where public hospitals indemnify medical practitioners, hospitals may face increasing exposure to liability for decisions made or influenced by AI. A healthcare provider confronting liability for an AI related failure in the standard of care is then left to consider options for recovery or contribution from the developer or supplier of the technology, which may be determined by the contractual arrangements in place.

While the law may identify duties of care within AI supported healthcare, a significant challenge lies in assessing a breach of that duty, particularly given the complex and opaque nature of AI decision-making. Determining a breach of duty in a medical negligence claim involves assessing whether a medical professional’s conduct fell below the standard of care expected of a reasonably skilled practitioner.⁷ AI complicates this assessment by raising questions about whether reliance on, or rejection of, an AI recommendation meets that reasonable standard and how peer professional opinion defences apply when AI norms are still evolving.⁸ For example, if a doctor relies on an AI diagnosis later found to be incorrect, the key issue is whether a reasonable doctor would have identified the error through additional scrutiny or by obtaining second opinions. If the AI’s error was so obscure that no reasonable clinician could have detected it, the clinician may not be deemed negligent.

These evaluations increasingly rely on expert input and evolving guidelines, which advocate for cautious and supplementary use of AI.⁹ There is growing recognition that new types of expert witnesses, with both clinical and AI literacy, may be needed to assess malpractice claims involving AI.¹⁰

'Black box' paradox

The inherent 'black box' nature of AI systems complicates the assessment of breaches, whereby the internal logic or reasoning behind an AI system’s output is unclear or inaccessible, making it difficult to trace errors or justify decisions. Claimants will be seeking to demonstrate that a clinician’s reliance on AI was unreasonable, which is difficult when the system’s decision-making processes are not transparent or fully understood, even by its developers.

This hurdle poses significant practical challenges, potentially requiring innovative legal strategies or burden of proof adjustments, as suggested internationally by the European Union.¹¹ Additionally, scenarios may arise where patient harm results solely from an AI failure, without clinician negligence, shifting liability toward manufacturers or developers under product liability.

Overall, the integration of AI into healthcare demands clearer professional guidelines and will likely require legal reform to address the evolving challenges of establishing breaches and assigning responsibility.

Professional standards

Back in August 2023, the Australian Medical Association issued a position statement supporting the use of AI in healthcare, provided it remains patient-centered and enhances the health and well-being of individuals and the broader community.¹² Building on this, the Australian Health Practitioner Regulation Agency and the National Boards released guiding principles to help health practitioners assess the use of AI in their practice.¹³

These principles reflect existing obligations within professional codes of conduct and urge practitioners to uphold responsibilities, including accountability, thorough understanding of the AI tool, transparency of use, informed consent and ethical and legal compliance. The consistent message from regulators is clear: AI is not a substitute for human expertise, but a valuable complement when applied thoughtfully and ethically.

Therapeutic goods administration guidance on AI scribes

On 30 January 2026, the Therapeutic Goods Administration (TGA) clarified that digital scribes,¹⁴ also referred to as AI scribes or ambient scribes, are regulated as medical devices under the Therapeutic Goods Act 1989 (Cth) when they perform functions beyond simple transcription.¹⁵ If a digital scribe is designed to analyse or interpret clinical conversations, such as generating diagnoses, treatment recommendations, or clinical insights not explicitly stated by the practitioner, it is considered to have a therapeutic purpose and must be included in the Australian Register of Therapeutic Goods before being supplied.¹⁶

These regulated products must meet all relevant safety, privacy and compliance obligations, including informed consent, review of software updates that may alter functionality and mechanisms for reporting safety concerns or non-compliance. The TGA also clarifies that digital scribes used solely for transcription, without interpretation or clinical decision support, do not fall under the medical device definition.¹⁷ As AI tools become more embedded in clinical practice, including cases where their use is essential in managing practitioner workloads, these guidelines underscore the growing need for legal clarity, patient consent and robust oversight.

AI in mental health

AI is also reshaping mental health care. A 2024 study published in JMIR Mental Health surveyed both community members and mental health professionals across Australia,¹⁸ revealing that 28% of community members and 43% of mental health professionals are already using AI tools — primarily for quick support, personal therapy, research and clinical documentation. Most respondents acknowledged the benefits of AI with respect to accessibility, cost reduction and efficiency, although nearly half of the respondents reported concerns surrounding privacy, ethical risks, reduced human connection and potential misuse.¹⁹

This study highlights the dual-edged nature of AI’s introduction into mental healthcare. Whilst AI offers scalable support for practitioners, it also presents new legal and ethical complexities that current frameworks are only beginning to address.

Preparing for the next phase of AI integration in healthcare

AI offers enormous potential benefits, but it also raises tough questions about safety, accountability and patient rights. Key questions which may arise for industry stakeholders, along with practical guidance on options to address these in the Australian healthcare context include the following.

Who should be responsible when AI is involved in clinical decisions?

AI use will not create a loophole in the duty of care. Clinicians and hospitals remain accountable at common law to patients for exercising reasonable care, skill and judgment while using AI tools. An option for hospitals or other healthcare institutions is to appoint a responsible officer to oversee AI use and ensure a human is always involved in key decisions, adhering to the global 'human-in-the-loop' standard for responsible AI use.²⁰

If a patient is harmed by AI, is the law ready?

The law is currently in its early stages of development. Legislative change is likely on the horizon and the impact may be significant. In Queensland, for instance, amending the Civil Liability Act 2003 (Qld), to introduce a rebuttable presumption of breach, similar to that suggested by the Law Council of Australia in their review of AI and the ACL. Introducing this rebuttable presumption would mean the court would presume a breach has occurred until evidence is presented to prove otherwise. This would essentially shift the burden to providers to prove non-negligence.²¹ Legislators and regulators may also look at ways to empower courts and healthcare providers, to demand transparency from AI developers where possible, to address the 'black box' problem.

What about training, are clinicians prepared to use AI safely?

Currently, the level of training clinicians receive is unregulated and variable. Regulators may in future establish standards for AI-specific training and certification, especially for high-risk tools like autonomous surgical robots. Courts could then consider compliance with these standards when evaluating negligence claims.

How do healthcare providers comply with their obligations in the meantime?

Medical practitioners should proceed cautiously and independently assess AI outputs with reference to their own clinical judgement before acting on them. Clear documentation, including of their independent clinical reasoning, will be important when using AI as a decision support tool. Practitioners should also be aware of the risks inherent in AI use such as bias and hallucinations.

Hospitals and other healthcare institutions should carefully vet AI tools before implementing them and introduce robust governance processes which include pathways for monitoring and reporting adverse events. Care must be taken to ensure compliance with data privacy standards with reference to the Office of the Australian Information Commissioner and to make sure timely software updates are implemented.

Conclusion

Whilst AI is a powerful tool enabling substantial progress in healthcare, its rapid adoption must be matched by the careful evolution of the legal principles that regulate its use. Ensuring that caution, accountability and patient protection remain at the forefront is essential to maintaining safety and trust as these technologies become increasingly embedded in clinical practice.

Footnotes

1. Department of Health, Disability and Ageing Safe and Responsible Artificial Intelligence in Health Care — Legislation and Regulation Review Final ReportReport (2025).

2. Above, at 4.

3. Above n 1, at 12.

4. Above n 1, at 5.

5. Competition and Consumer Act 2010 (Cth), Sch 2 (Australian Consumer Law).

6. P Nolan and R Matulionyte 'Artificial Intelligence in Medicine: Issues When Determining Negligence' (2023) 30(3) Journal of Law and Medicine593–615.

7. Civil Liability Act 2003 (Qld), s 22. Civil Liability Act 2002 (NSW), s 5O. Wrongs Act 1958 (Vic), s 59. Civil Liability Act 1936 (SA), s 41. Civil Liability Act 2002 (WA), s 5PB. Civil Liability Act 2002 (Tas), s 22.

8. Above.

9. Above n 1.

10. C Terranova and others 'AI and professional liability assessment in healthcare. A revolution in legal medicine?' (2024) 10 Frontiers in Medicine.

11. European Parliament Legislative Observatory, Adapting non-contractual civil liability rules to artificial intelligence (AI Liability Directive), 2022, accessed 15 April 2026.

12. Australian Medical Association Position Statement: Artificial Intelligence in HealthcareReport (2023).

13. Australian Health Practitioner Regulation Agency & National Boards, Meeting your professional obligations when using Artificial Intelligence in healthcare, August 2024, accessed 15 April 2026.

14. Therapeutic Goods Administration, Digital Scribes, January 2026, accessed 15 April 2026.

15. Therapeutic Goods Act 1989 (Cth), s 41BD.

16.  Above n 14.

17. Above n 1.

18. S Cross, I Bell, J Nicholas, L Valentine and others 'Use of AI in Mental Health Care: Community and Mental Health Professionals Survey' (2024) 11 JMIR Mental Health 6.

19. Above n 11, at 6.

20. S Bakken 'AI in health: keeping the human in the loop' (2023) 30(7) Journal of the American Medical Informatics Association 1225–26.

21. Law Council of Australia Review of AI and the Australian Consumer Law Report (2024) 84.