Earning Trust for AI in Health 2025

Proposed framework for regulating AI in health FIGURE 1 Regulatory provisions setting the guardrails Regulatory provisions need well-structured guidelines to have real-world effects, which can be developed using industry standards and experience Adapt post-market surveillance and monitoring Ensure market access guidelines and protocols consider the capabilities of AI technologies to/uni00A0evolve post-deployment Transform guidelines and regulations into actionable procedures To support/uni00A0the private sector to ensure real-world compliance and effectivenessRegulatory provisionsIndependent testing Provide non-binding guidance for all stakeholders Under the responsibility of governments or dedicated regulatory bodies, supported by independent expertise, notably from the innovation and academic communitiesIndependent guidelines setting Operationalization Source: World Economic Forum and Boston Consulting Group analysis Private-sector involvement should be carefully designed to preserve regulatory integrity and independence, while taking advantage of the sector’s unique skills and capabilities. Thus, it is essential to mobilize the private sector at the right steps of the regulatory process (see Figure 1): –First, the private sector should be consulted in the upstream phases of the regulatory and guidelines development processes. The private sector can support the ecosystem to provide non-binding guidance that over time will inform legislation on AI in health. –Second, private-sector involvement should extend to the translational aspects of legislation. A legislative framework sets out a high-level vision for the roles of AI technologies in society, paired with appropriate boundaries and guardrails. The development and implementation of guidelines that aim to realize this high-level vision can benefit greatly from industry input, offering insights into how that vision can be realized through purposeful and public value-driven innovation. For instance, the world’s first international standard dedicated to AI management systems (ISO/IEC 42001:2023) was developed through international collaboration involving diverse stakeholders.31 –Third, the private sector is ideally placed to develop and scale pre- and post-market testing and monitoring approaches to detect deviations in the performance of AI technologies and correct them. The private sector is best positioned to provide the technical expertise needed to build real- time monitoring capabilities. For example, US company Galileo has developed a platform that embeds accurate evaluations directly into AI development workflows.32 Regulators already create frameworks for private-sector engagement. However, most of the companies interviewed for this paper reported challenges in making consistent and meaningful contributions. Appropriately involving private-sector actors in the policy process and implementing feedback loops can help ensure that guidelines for AI in health keep pace with technological advances. Earning Trust for AI in Health: A Collaborative Path Forward 14