AI in health care system: ‘Innovation cannot outrun accountability’, says Omprakash Subbarao

Bengaluru: As India’s digital health mission matures beyond appointment-booking apps and teleconsultations, a quieter but more consequential shift is underway starting from access to accuracy. 

The next phase is not about scaling downloads, but about validating algorithms. With AI poised to read X-rays, interpret biomedical signals and guide clinical decisions at scale, questions of safety, transparency and scientific rigour are moving to the forefront. 

A part of this transition is Omprakash Subbarao, chief of CORE Labs at IISc and advisor to National Health Authority, helping shape how AI enters India’s public health system not as a black box, but as a tool that must earn trust before it earns reach. 

In an interview to Amullya Shivashankar of DH, Subbarao stressed the scientific validation, safety frameworks and regulatory guardrails for AI before population-scale deployment. 

What are the real public-health risks of deploying opaque AI models at national scale? 

Along with opacity, accountability of opacity makes AI risky. Large health systems, regardless of whether they are human or AI-based, can be risky if their underlying decision processes cannot be scrutinised. Implementing AI through models including appropriate levels of transparency, accountability and human oversight is crucial. It should be decision-supporting, not decision-replacing system. 

How far are today’s biomedical signals from being clinically reliable in Indian conditions?

The progress is more than what is recognised. The diversity of India, be it, psychologic or environmental and others suggests that worldwide datasets may not have captured the true reality. Indian research organisations play an important role, here, due to their access to empirical evidence that apply directly to Indian context as an important precursor to translating biomedical signals into dependable clinical applications. 

Is India’s regulatory system capable of evaluating AI safety beyond accuracy metrics?

As India is in the process of changing its regulatory system, the importance will be placed on safety, bias, functioning of AI and its monitoring systems. The greatest strength lies in adoption rather than pausing innovation through an inflexible regulatory framework at an early stage. 

If India gets AI in healthcare wrong now, what long-term damage could it do?

Public trust. Approaching healthcare AI with humility and accountability will actively build public trust, advancing innovation and improving care, but transparency is the key. Achieving correctness does not involve merely avoiding making errors, but acknowledging limits, correcting one’s direction when needed and allowing humans to be part of the system. 

Who should be accountable when a “black box” medical AI fails?

The accountability has to be shared across the table including clinicians, regulators, institutions and systems of care. When responsibility is shared across the board such as developers, implementers and public authorities, it creates a clearer pathway for innovation with responsibility. This being a collaborative effort will focus towards shared responsibility and safeguards patient interests while technological progress continues at scale.