Earning Trust for AI in Health 2025

Examples of private-sector involvement in the regulatory process FIGURE 2 Renewed private-sector engagement: Regulatory provisions need well-structured guidelines to have real-world effects, which can be developed using industry standards and experience Private-sector consultation to provide expert advice and practical insights2024 consultation on general-purpose AI models PPP to industrialize post-market surveillance and monitoring of the increasing number of AI applications Private-sector collaboration to transform guidelines and legislation into actionable proceduresRegulatory provisions OperationalizationIndependent testingIndependent guidelines setting 2023 consultation on AI regulation The “Validate” programme to evaluate bias and accuracy of AI modelsPlatform to test, monitor and deploy AI systems at scaleEvaluation laboratory to assess AI, including its ethics Collaboration to harmonize standards and reporting for AICollaboration to establish best practices for deploying AI in health Professional association to develop standards, including for AI Global non-profit aiming at building responsible AI solutions in health Source: EU consultation: https://digital-strategy.ec.europa.eu/en/consultations/ai-act-have-your-say-trustworthy-general-purpose-ai; UK consultation: https://www. gov.uk/government/consultations/ai-regulation-a-pro-innovation-approach-policy-proposals; FDA consultation: https://www.fda.gov/media/122535/download; Mayo Clinic Platform: https://www.chiefhealthcareexecutive.com/view/ai-success-in-healthcare-requires-transparency-public-private-partnership Quality assurance resources are being established to evaluate and validate AI models independently, using consensus-driven standards and best practices. These resources are structured environments, often in form of labs hosted at a network of quality assurance resource providers (QARPs). They can use a set of community-approved best practices for developing trustworthy health AI, such as those proposed by the Coalition for Health AI (CHAI) or the US National Academy of Medicine’s AI Code of Conduct. Beyond model evaluation, such assurance resources in a network of QARPs can serve as a key infrastructure investment across an AI model’s entire life cycle (development, deployment, post-deployment governance and monitoring), supporting a range of critical stakeholders in the health AI ecosystem. For example, they can accelerate model training given their access to robust, heterogeneous data, speeding up development and improving model performance across communities, or they can support longitudinal governance for deployed AI models. The role of QARPs and assurance resources continues to evolve and expand as the concept is tested and scaled. At the end of 2024, CHAI introduced a framework to certify quality assurance resources primarily led by the private sector. Similarly, in the EU, a network of testing and experimentation facilities (TEFs) is being established33 – hospital platforms, living labs and laboratory testing facilities, for example. These facilities will give innovators the capacity to carry out tests and experiments on their AI technologies in large-scale and sustainable real or realistic environments.3.3 Quality assurance resources: An approach to PPPs for independent testing and training Earning Trust for AI in Health: A Collaborative Path Forward 15