Knowledge, Attitudes and Perceptions of the Regulation of Medical Artificial intelligence in South Africa: A survey of regulatory professionals

Abstract

The integration of Artificial Intelligence (AI) into healthcare, particularly in medical devices, presents significant opportunities for improving diagnostics, monitoring, and treatment. However, these innovative technologies also pose complex regulatory challenges. In South Africa, regulatory professionals play a pivotal role in overseeing the safe and ethical application of medical AI, yet the regulatory framework remains underdeveloped. This study explores the knowledge, attitudes, and perceptions of South African regulatory professionals regarding medical AI regulation and identifies gaps in existing frameworks. Aim: This study aims to evaluate the level of knowledge, understanding, and perceptions of medical AI among regulatory professionals in South Africa. Findings could inform further investigations aimed at developing recommendation strategies to improve regulatory frameworks within South Africa. Methods: A quantitative, cross-sectional survey was conducted with 106 regulatory professionals recruited through the South African Medical Technology Industry Association (SAMED) and LinkedIn, yielding a 77.94% response rate. The questionnaire assessed AI knowledge, understanding of AI applications, regulatory challenges, and attitudes toward AI in healthcare. Data were analysed using descriptive and inferential statistical methods, including Kruskal-Wallis (H = 27.86, p < 0.001), Mann-Whitney U, and Chi-square tests (χ² = 38.72, p < 0.001). Likert-scale responses measuring attitudes and perceptions were analysed for central tendency (mean: 3.79, median: 4.49) and variability (standard deviation: 0.91–1.14). Results: Regarding knowledge, 99.06% of respondents were familiar with artificial intelligence (AI), with 86.8% able to correctly define machine learning (Mann-Whitney U = 125, p < 0.001). However, only 46.2% were aware of AI applications in healthcare (Mann-Whitney U = 450, p = 0.005), and an even smaller percentage (23.6%) reported the use of medical AI in their organizations (Mann- Whitney U = 600, p = 0.09). In terms of regulatory gaps, 81% of respondents felt that South Africa lacked adequate regulatory frameworks for medical AI (Chi-square test: χ² = 95.79, p < 0.001). When considering perceived challenges, safety and efficacy emerged as the most critical concerns, identified by 68.9% of respondents, followed by data privacy (63.2%) and algorithm transparency (50.9%). Additionally, 52.8% highlighted the importance of continuous learning in AI systems, and 56.6% emphasized the need for international harmonization of regulations. Attitudes xi toward AI were generally positive, as revealed by Likert-scale analysis. There was strong agreement with statements highlighting AI’s potential, such as “AI has great potential for use in medical devices for diagnosis, prognosis, and treatment” (mean: 4.49, SD: 0.91). However, scepticism persisted, with 39.6% of respondents neutral on whether AI improves patient outcomes, and the statement “AI will replace physicians in the future” receiving the lowest agreement (mean: 2.12, SD: 1.14). Finally, the demographic profile of respondents showed that the majority were female (72.6%) and highly educated, with 60.4% holding postgraduate qualifications. Conclusion: The study highlights a substantial gap between general AI knowledge and practical regulatory understanding among South African professionals. Statistical tests underscored significant differences in knowledge and attitudes. Respondents emphasised the need for robust, localized regulatory frameworks addressing safety, efficacy, data privacy, and ethical concerns. International collaboration with health authorities and harmonized standards were identified as critical for fostering innovation and ensuring the safe integration of AI into the South African regulatory framework. Implications: Targeted education, capacity building, and regulatory reforms are recommended to enhance oversight of the integration and implementation of AI in medical devices and diagnostics in South Africa.