The Future of AI Enabled Health 2025

On the second point, it must be acknowledged that the rapid pace of AI development necessitates a new approach to expanding validation capacity. This can be achieved by delegating the validation process, under the supervision of regulators and governments, to involve not only government bodies but also non-profits, care providers and private-sector players. Unlike the slower pace of drug and medical device development, AI software evolves rapidly and public capabilities alone cannot keep pace. New testing and validation methods are required, and private-sector expertise is crucial in developing these. While the private sector cannot directly regulate AI, its collaboration helps to ensure that new regulatory frameworks are well informed, practical and agile enough to keep up with technological advances. Yet the private sector is a diverse landscape, and large organizations are more likely to be able to free up resources to participate in discussions on guidelines and standards. The academic world might suffer from the same lack of resources as smaller private-sector organizations. This is why public–private partnerships, including funded PPPs as defined by the American National Standards Institute (ANSI),20 should be considered for long- term sustainability, as well as the inclusion of small and medium-sized enterprises (SMEs) or academics in the design and rollout of standards, guidance and, eventually, regulations. Collaboration initiatives are also promising. For example, the Coalition for Health AI (CHAI) brings together a diverse array of stakeholders to drive the development, evaluation and appropriate use of AI in healthcare. CHAI has developed a certification framework to establish a network of quality assurance laboratories that evaluate AI models for healthcare use. AI regulation often follows a one-size-fits-all approach, which is inadequate for the diverse and rapidly evolving nature of AI technologies. The nature of genAI, which is non-deterministic and can evolve as data is collected during use, requires more flexible and nuanced regulatory approaches compared to traditional AI. Current regulatory frameworks struggle to adapt to these technologies, as traditional methods are ill-equipped to manage their complexities and rapid evolution. GenAI’s unique characteristics and risks demand a regulatory framework that is both adaptive and forward-thinking, ensuring that regulations keep pace with technological advances. A stronger focus on post-market surveillance could be considered as a way of detecting new risks early on, address errors and biases and adapt iteratively. Finally, data protectionism hampers innovation and limits the potential for AI advances. It restricts the ability to develop robust AI using unbiased datasets and to validate AI tools in local contexts. To facilitate global adoption and development, it is essential to ensure the convergence of data models and exchange standards; for example, through locally controlled but globally federated datasets.21 These datasets enable AI solutions to be developed and validated for different local populations, ensuring greater accuracy and safety while preserving privacy. Difficulty in building trust within a complex ecosystem A global study found 44% of people surveyed expressed a willingness to trust AI in health applications,22 reflecting a cautious optimism about its potential benefits and concerns about its implementation and oversight. This cautious attitude is supported by data showing that 67% of health leaders in the US trusted AI technology to process medical records by 2020, a significant increase from 54% in 2018.23 However, the acceptance of AI in health systems remains at risk due to broader concerns about misinformation and the quality of health information. This sentiment is echoed in consumer attitudes to AI in different countries, as illustrated in Figure 7, where feelings about AI are mixed, with more than 40% expressing concern in the US, Switzerland, the United Kingdom, France and Australia, while fewer than 20% share this concern in China, India, Thailand, Saudi Arabia, Indonesia and Mexico. Building trust in AI for health requires a concerted effort on both the regulatory and business fronts. Transparency is a cornerstone in this endeavour. Regulatory bodies must ensure openness, clear communication and full disclosure of important facts about AI technologies to alleviate public concerns. As emphasized by the WHO: “it is fundamental to consider streamlining the oversight process for AI regulation through […] engagement and collaboration [among key stakeholders]”.24 Equally important is integrity, with regulations enforcing consistent honesty and ethical behaviour, ensuring that actions align with stated goals. Human oversight also plays a crucial role, especially given the challenges with genAI, such as hallucinations. Ensuring that humans remain involved in validating AI outputs and supporting decision-making is essential for maintaining trust and efficacy. Furthermore, avoiding the anthropomorphization of AI is vital, as this creates confusion between the perception of capabilities and the limitations of AI. Humans must remain accountable to ensure trust, safeguard efficacy and address potential issues in AI systems. AI ambassadors can play an important role in communicating the benefits and limitations of AI- based products, helping to build a clear strategy for trust and transparency. From a business perspective, demonstrating integrity, competence and potential is fundamental. The Future of AI-Enabled Health: Leading the Way 21