The draft AI policy: Implications for healthcare providers, insurers and intermediaries

Artificial intelligence (AI) is increasingly being integrated into South Africa’s healthcare and insurance sectors. Examples include systems that analyse clinical data, assist with diagnosis, support underwriting decisions, and detect fraud. While these tools enhance efficiency and decision-making, they do not replace professional judgment.

South Africa has not had a dedicated legal framework regulating AI. Instead, existing legal principles relating to professional accountability, negligence, confidentiality and data protection have applied indirectly. These frameworks were not designed with AI in mind. This position is now evolving.

On 10 April 2026, the Department of Communications and Digital Technologies published the Draft South Africa National Artificial Intelligence Policy (the Draft AI Policy) for public comment. The Draft AI Policy sets out a proposed framework for the governance of AI across all sectors, including healthcare and insurance. Its stated aim is to ensure that AI is used safely, ethically and responsibly, while promoting innovation and protecting constitutional rights.

This article outlines the proposed implementation of the Draft AI Policy and highlights key implications for healthcare practitioners, healthcare facilities, insurers, underwriters, brokers and insured parties.The Draft Policy does not introduce immediate legal obligations. Instead, it proposes a phased implementation over approximately three years, with early focus on high-risk sectors such as healthcare and insurance. The three phased approach entails the following:

Phase 1 (2025/26)

• Finalisation of the national AI policy
• Identification of prohibited and high-risk AI use cases
• Development of national AI guidelines

Phase 2 (2026/27)

• Introduction of regulatory requirements for high-risk AI systems
• Development of sector-specific strategies, including healthcare and insurance
• Publication of draft rules for medium- and low-risk AI systems

Phase 3 (2027/28)

• Full implementation of the regulatory framework
• Establishment of oversight mechanisms and institutions
• Alignment of existing legislation with AI developments

While the phased approach avoids immediate disruption, it signals that regulatory requirements will progressively tighten, particularly in high-risk environments.

The Draft Policy emphasises “human-in-the-loop” decision-making in high-risk contexts. Healthcare practitioners remain ultimately responsible for clinical decisions, even where AI tools are used. It also introduces a requirement for explainability. Healthcare practitioners must understand how AI informs clinical decisions and be able to explain this to patients in clear terms. This has implications for informed consent and standard of care.

The use of AI must also align with constitutional rights, including dignity, bodily integrity and protection from harm.

Policy requirements

Healthcare facilities using AI for diagnosis, treatment planning or predictive analytics are likely to fall within the high-risk category. Facilities may be required to:

• conduct AI risk assessments and audits
• ensure compliance with data protection laws, including POPIA
• implement appropriate governance and oversight structures
• obtain certification for certain high-risk AI systemsCross-border data flows and cloud-based AI solutions may attract additional regulatory scrutiny.

The Draft AI Policy has direct implications for insurers using AI in underwriting, pricing, claims assessment and fraud detection. Key requirements include:

• transparency and explainability of AI-driven decisions
• the ability for affected persons to challenge such decisions
• testing for bias and discrimination
• human oversight in materially impactful decisions

The Draft Policy reinforces existing obligations under section 71 of POPIA relating to automated decision-making. It also proposes the establishment of an AI Insurance Superfund, which may compensate individuals harmed by AI-driven outcomes where liability is uncertain. This may have cost and structural implications for the insurance industry.

What it means for the advisory industry

Brokers and intermediaries will be expected to understand AI-related risks, explain these risks to clients and ensure appropriate disclosure and advice. This raises the standard of advisory competence, particularly where AI is integrated into products or decision-making processes.

Underwriters using AI systems will similarly remain accountable for decisions, notwithstanding reliance on automated tools.The Draft Policy strengthens the rights of individuals affected by AI systems. These include the right to be informed of AI-driven decisions and challenge such decisions, as well as the right to enhanced data protection safeguards.

Where harm arises from AI-related outcomes, affected persons may have access to compensation mechanisms, including the proposed AI Insurance Superfund. The Draft AI Policy signals a significant shift in the regulation of artificial intelligence in South Africa. Healthcare and insurance are identified as high-risk sectors, with increased emphasis on accountability, human oversight and protection of constitutional rights.

While the regulatory framework will be implemented progressively, stakeholders should begin preparing for increased scrutiny and evolving compliance requirements. Members of the public and interested stakeholders are invited to submit written comments on the Draft AI Policy by 10 June 2026 at 16h00.

A copy of the Draft AI Policy may be accessed by clicking here.