What sets AI systems and AI-powered medical robots apart from all other forms of advanced medical technology is their ability to operate at least to some degree autonomously from the human health care practitioner and to use machine-learning to generate new, often unforeseen, analysis and predictions. This poses challenges under the current framework of laws, regulations, and ethical guidelines applicable to health care in South Africa. The article outlines these challenges and sets out guiding principles for a normative framework to regulate the use of AI in health care. The article examines three key areas for legal reform in relation to AI in health care. First, it proposes that the regulatory framework for the oversight of software as a medical device needs to be updated to develop frameworks for adequately regulating the use of such new technologies. Secondly, it argues that the present HPCSA guidelines for health care practitioners in South Africa adopt an unduly restrictive approach centred in the outmoded semantics of telemedicine. This may discourage technological innovation that could improve access to health care for all, and as such the guidelines are inconsistent with the national digital health strategy. Thirdly, it examines the common law principles of fault-based liability for medical negligence, which could prove inadequate to provide patients and users of new technologies with redress for harm where fault cannot clearly be attributed to the healthcare practitioner. It argues that consideration should be given to developing a statutory scheme for strict liability, together with mandatory insurance, and appropriate reform of product liability pertaining to technology developers and manufacturers. These legal reforms should not be undertaken without also developing a coherent, human-rights centred policy framework for the ethical use of AI, robotics, and related technologies in health care in South Africa.
Keywords: Artificial intelligence; ethics; health care; health policies; machine learning.