Artificial Intelligence has been changing the lives of people in India in many ways, whether in educational institutions or in the work environment, although most of the time, people may not be aware of that. However, despite its rapid application, regulatory frameworks exist but remain not fully adapted to AI-driven drug discovery. The release of the February 2026 India Artificial Intelligence Governance Guidelines, which include the Secure Artificial Intelligence in Health Initiative (SAHI) and the Benchmarking Open Data Platform for Health Artificial Intelligence (BODH), these initiatives signal a shift away from a purely light-touch approach toward more structured oversight of AI in healthcare.

However, the current “light touch” regulatory regime that has been very successful in sectors like fintech and digital infrastructure leaves critical gaps in how AI used in drug discovery—particularly in preclinical stages—is governed. While the Medical Devices Rules (2017), New Drugs and Clinical Trials Rules (2019), and Digital Personal Data Protection Act (2023) govern the use of AI devices and AI generated data during the course of research, and as of 2026 there are limited statutes that makes it mandatory to disclose the workings of an algorithm used in designing drugs using artificial intelligence. India’s pharma industry, which had earned a reputation as the “pharmacy of the developing world,” could lose far too much otherwise.

Pillar I: Closing the SaMD Gap with “Glass-Box” AI

AI applications that deal with patients at present – diagnostics, imaging, decision-support – are strictly governed and audited, validated and integrated as Software as a Medical Device (SaMD).

However, the AI applications used to design the very drugs used in those hospitals? While existing frameworks indirectly apply, there is no AI-specific oversight for preclinical drug discovery models. This remains a critical blind spot.

Modern drug discovery cannot happen without prediction modeling, especially graph neural networks, which sift through thousands of molecules for toxicity and efficiency, deciding which molecules go into research labs, which go into clinical trials bypassing animals ones, and which get delivered to the patient. Despite this, the governing agencies consider them to be "lab tools" rather than actual decision-making software.

Such governance is unacceptable moving forward. With AI firmly entrenched in the R&D process, there can be no room for errors, and India's potential for attracting foreign investment could be squandered if this loophole is not sealed now. It could invite a potential new horde of malicious actors. Looking ahead, India should legislate the requirement of a glass-box assessment for preclinical AI models with clearly defined interpretability and validation standards. This can be done by a CDSCO-led task force, combining regulators, biotech firms, and computational chemists. Innovation thrives only when accountability is baked into the process.

Pillar II: Building a Global South Data Ledger

Transparency alone will not save broken AI. Models will only be as good as the data they use to train. In today’s world, the most extensive datasets for toxicology and bioactivity analysis come primarily from Western nations. Using such data without any adjustments will likely perpetuate existing biases that fail to reflect the Indian population’s genetic makeup, environmental exposure, and healthcare situation. Initiatives such as the BODH platform and the Ayushman Bharat digital health ecosystem represent early efforts to address this imbalance but are not the complete solution.

While these efforts are still evolving, India cannot afford to wait for perfect and customized datasets to begin its work in the field. What needs to be done is building while adjusting. For example, India should initiate the development of a Global South Data Ledger. It would be an open-source, blockchain-based consortium where countries from the Global South can share their data on clinical, chemical, and toxicological aspects of healthcare.

Blockchain technology is much more than just a buzzword here; it helps create unalterable records, thereby guaranteeing data integrity regardless of national boundaries. The decentralized nature of blockchain ensures that no one player controls the data set, while smart contracts could ensure that privacy and sharing policies are adhered to.

This blockchain could integrate both public actors—such as governments from participating countries—and private stakeholders, creating incentives to share data that simultaneously strengthen national healthcare systems and attract foreign investment. This way, India could continue developing its biotechnology industry and gradually eliminate the existing biases.

Pillar III: Cyber-Biosecurity for Dual-Use AI

There is a hard truth at the heart of generative chemistry: the same AI that predicts toxicity can be repurposed to create it. This risk has been demonstrated in controlled experiments, including work by researchers such as Fabio Urbina, where drug discovery models were repurposed to generate toxic compounds. Most discussions frame AI as a benevolent tool—reducing animal testing, making safer drugs—but they overlook its dual-use potential.

The analogy is clear: in the early days of computing, malware warnings were everywhere. Today, the pop-ups are gone, but the security systems they enabled remain indispensable. AI in pharma needs the same principle.

India must implement a “cyber-biosecurity”—an emerging concept in security and policy circles addressing the intersection of biotechnology and digital threats— framework that treats generative chemistry risks as seriously as cybercrime. Mandatory internal screening layers should flag when models are generating harmful compounds, open-source developers must comply with safeguards, and deliberate circumvention should carry stiff legal penalties. This is not regulation for the sake of control—it is infrastructure for safe growth.

Conclusion: Leadership Through Accountability

AI is already beginning to change how new drugs are made. The question is whether India will be a leader or a follower. General ethics policies are not enough when the models themselves decide on the safety of a drug even before the molecule is created.

With transparency through AI in preclinical studies, a sovereign data network, and dual use considerations through cyber-biosecurity, India can secure its public health, investments, and position in the international market.

It is not just about minimizing risk. It is also an opportunity to make an international benchmark in intelligent and sovereign pharma that does not put people at risk in the name of speed.