The deployment of AI solutions in healthcare will require a “life cycle” approach, starting from defining the problem to collection, storage and management of data, verification and validation, and real-world performance, according to a policy framework for the sector released by the Union Health Ministry on Tuesday at the AI Impact summit.

“Once you develop an AI solution, how do you buy it, how do you monetise it? There is no way to determine that in the current mechanisms and procurement methods,” National Health Authority CEO Sunil Kumar Barnwal said.

Elaborating on the need for such a framework, he said, “One of the biggest challenges with AI is that it learns while it is being used but it can also drift sometimes. These challenges can be addressed only when we look at the whole life cycle of AI from data collection to training to deployment, continuous monitoring, and decommissioning if necessary.”

Barnwal headed the committee that developed the framework called Strategy for Artificial Intelligence in Healthcare for India (SAHI). This is the larger framework for using AI in healthcare, with an implementation roadmap that lists out which department does what currently in the works, he said.

The framework states: “AI systems in healthcare differ fundamentally from conventional digital tools: they influence clinical judgment, shape care pathways, and inform population-level decisions, often evolving through updates or learning mechanisms. As a result, AI adoption in healthcare cannot be treated as a one-time approval or deployment event.”

According to Barnwal, any AI solution in healthcare will have to follow principles that are already a part of the Ayushman Bharat Digital Mission (ABDM). Data privacy will have to be built in by design into any solution, he said, adding that patient consent would be required for sharing of data.

“The data under ABDM resides at the source, be it a hospital, laboratory or pharmacy. It doesn’t migrate to a central database. And, the patient has to give consent every time when the data moves from one place to another,” he said.

Training of health workers is also key to effective deployment of any AI solution, Barnwal said.

Union Health Minister J P Nadda said India laid the foundation for AI years ago. “We all know that AI does not operate in isolation. AI lives, breathes, thrives on the digital infrastructure. Recognising this early, India began laying the foundation almost a decade before. In 2015, we launched digital India with a clear objective to make India into a digitally empowered society and knowledge economy,” he said.

He said that the ABDM created the digital public architecture, without which AI could not be developed responsibly or at scale. Nadda also launched a system that will allow innovators to train their healthcare AI models on real-life data made available on a single platform.

Dr Catharina Boehme, from WHO-SEARO, said: “This is not simply a technology roadmap, it’s a public health strategy built to strengthen care, improve dec­i­sions, extend reach. It supports progress towards universal hea­lt­h­care and sustainable development goals… India has become the first country in Southeast Asia region to adopt such a comprehensive strategy and one of the first countries globally to have it. India has set a benchmark.”