A woman in a village cough into a mobile phone held by a health worker. Within a minute, an app tells the worker whether she should be sent for a tuberculosis test. Ten years ago, this would have sounded like science fiction. Today it is a pilot running in 91 districts across 17 States, part of India’s campaign to end TB.
This small scene says a great deal about artificial intelligence in the life sciences. The big promises make headlines. The real change is happening quietly, in clinics, laboratories and farms. Life science is the study of living things, from a single cell to a whole forest. It covers medicine, biology, pharmacy, agriculture, genetics and environmental science. For most of history, progress in these fields came from patient observation and years of trial and error. What has changed is the amount of data. One human genome holds about three billion letters of genetic code. A large hospital produces millions of medical images every year. Satellites photograph every farm in the country many times a month. No team of humans can read all of this. Machines can.
What AI actually does
Take away the jargon and AI in biology does one thing well. It finds patterns in large amounts of data and uses those patterns to make predictions. Show a computer thousands of chest X-rays already marked by expert radiologists, and it learns to spot the shadows that suggest disease. Give it the known shapes of proteins, and it learns to guess the shapes of proteins no scientist has yet seen.
That second example changed biology. Proteins are the working machines inside every cell, and what they do depends on how they fold into three-dimensional shapes. Working out the structure of a single protein once took a doctoral student several years of laboratory work. Around 2020, a program called AlphaFold, built by Google DeepMind, began predicting these shapes with accuracy close to laboratory methods. Its creators, Demis Hassabis and John Jumper, shared the 2024 Nobel Prize in Chemistry with David Baker, who uses computers to design new proteins from scratch. The database of predicted structures now covers more than 200 million proteins. It is free to use, which means a research student in Patna or Srinagar has the same starting point as a scientist at Cambridge.
The search for new medicines
Finding a new drug is slow and expensive. A medicine usually takes 10 to 15 years to travel from the laboratory to the pharmacy shelf, and industry estimates place the full cost at around 2.6 billion US dollars. Most candidates fail along the way. AI companies promised to change these numbers, and investors listened. Roughly 60 billion dollars has gone into AI-based drug discovery since 2019. Industry trackers counted more than 173 AI-designed drug programmes in clinical trials in early 2026, up from about two dozen at the end of 2023.
A word of caution belongs here. As of mid-2026, no drug discovered or designed by AI had received full approval from the US Food and Drug Administration. AI has made the early steps faster and cheaper, such as choosing a disease target and screening millions of molecules on a computer. The later steps, where a drug is tested on thousands of patients, remain as difficult as before. An estimate by the Boston Consulting Group suggests AI roughly doubles a drug’s chance of reaching the market, from about 5 to 10 per cent to somewhere between 9 and 18 per cent. That is a real gain. It is no miracle. The first group of AI-designed drugs has now reached the final stage of human trials, and their results, expected around the end of this year, will tell us a lot.
For India, often called the pharmacy of the world for its supply of affordable generic medicines, this moment matters. AI gives Indian companies and universities a chance to move up from copying molecules to creating them. The best place to start is with diseases that large global firms tend to ignore because poor patients cannot pay high prices. Tuberculosis, dengue, kala-azar and drug-resistant infections fall in this group.
There is another opening that is close to my own institution. Jamia Hamdard was founded on the vision of Hakeem Abdul Hameed Sahab, who believed the knowledge of Unani medicine and the methods of modern pharmaceutical science should work together. Classical texts in Unani and Ayurveda describe hundreds of plants and preparations. AI can study the chemical compounds in these plants, predict which ones may act on a known disease target and point scientists toward the most promising leads. Every such lead must still pass through proper laboratory tests and clinical trials. Used this way, AI can help our traditions speak in the language of evidence.
In the clinic and the community
India’s health system faces a basic shortage. There are too few radiologists, pathologists and specialists for a country of 1.4 billion people, and most of them work in cities. This is where AI has already started to help.
The Indian Council of Medical Research and the Institute for Plasma Research have jointly built a tool called DeepCXR. It reads chest X-rays, sorts them into normal and abnormal, and flags signs of TB and other lung diseases so that a doctor can confirm them. The tool is given free of cost under the National TB Elimination Programme. The government told Parliament last year that it had been deployed in eight states and Union Territories, and that handheld X-ray machines with AI support were used to screen vulnerable people during the 100-day TB Mukt Bharat campaign.
Scientists at ICMR’s National Institute for Research in Tuberculosis in Chennai tested another AI system that reads the laboratory strips used to detect drug-resistant TB. Across ten laboratories, it matched expert readings with accuracy between 92 and 100 per cent. AIIMS has developed an AI-assisted camera that checks the retina for damage caused by diabetes, a leading cause of preventable blindness. The national telemedicine service, e-Sanjeevani, uses AI to suggest possible diagnoses to the doctor at the other end of the call.
In each of these cases the machine sorts, flags and suggests, while a trained person makes the final decision. That division of work is sensible, and it should stay that way. A district hospital with one radiologist can use AI to push the most worrying X-rays to the top of the pile. The doctor’s skill is spent where it counts.
Fields, forests and food
The life sciences reach well beyond hospitals. In May 2025, India released two genome-edited rice varieties that scientists say give higher yields and need less water. Plant breeders now use AI to decide which genes to change and to predict how a new variety will behave in different soils and climates. This cuts years from breeding programmes at a time when heat waves and erratic monsoons are already hurting harvests.
On the ground, a farmer can photograph a sick leaf and get a likely diagnosis of the pest or disease on her phone. Satellite images analysed by machine learning help forecast crop yields and warn of drought. Ecologists use AI to identify birds and animals from their calls recorded in forests, which makes it far easier to count species and protect them.
India’s policy push
The government has noticed. In August 2024, the Union Cabinet approved the BioE3 policy, short for Biotechnology for Economy, Environment and Employment. It is India’s first dedicated national policy for biotechnology. It sets up a network of Bio-AI hubs, biofoundries and biomanufacturing centres, many in partnership with private industry. In August 2025, the Department of Biotechnology and the Ministry of Electronics and Information Technology signed an agreement under which the IT ministry provides computing power and data infrastructure and the biotechnology department contributes biological datasets.
The ambition is large. India’s bioeconomy grew from about 10 billion dollars in 2014 to roughly 165.7 billion dollars in 2024, and the official target is 300 billion dollars by 2030. The IndiaAI Mission, meanwhile, is building shared computing capacity that researchers in smaller institutions could never afford on their own. The Genome India project has sequenced 10,000 genomes from across the country’s communities, giving scientists data that reflects Indian populations.
Where implementation succeeds or fails
Good policies are a beginning. Whether AI improves lives in India will depend on how carefully it is put into practice, and several problems need attention.
The first is data. An AI system learns from the examples it is given. If those examples come mostly from patients in Europe and North America, the system may perform poorly on Indian patients, whose genetics, diets and disease patterns differ. A skin-disease model trained mostly on fair skin can miss conditions on darker skin. India must build large, well-labelled datasets of its own, and projects like Genome India are a good start.
Privacy comes next. Health records and genetic data are among the most personal information a person has. The Digital Personal Data Protection Act of 2023 gives citizens rights over their data, and ICMR issued ethical guidelines for AI in biomedical research and healthcare in 2023. These rules must be followed in spirit. Patients should know when their data is used and should be able to say no.
Then there is testing. A tool that works well in a Bengaluru laboratory may stumble in a crowded primary health centre with poor lighting and old equipment. Every AI system meant for patients should be tested in real Indian settings before it is rolled out widely, and checked again after deployment. ICMR’s health technology assessment of AI tools for TB screening, which found that they increased detection while saving costs, is the kind of evidence we need more of. Our drug regulator also needs the staff and expertise to evaluate AI-based medical products quickly and fairly.
Basic infrastructure is often forgotten. An AI tool is of little use in a clinic where the power fails every afternoon or the internet connection drops. Investment in AI has to go hand in hand with investment in electricity, connectivity and maintenance.
The biggest gap may be people. India needs biologists who can write code and computer scientists who understand cells and molecules. Our universities still keep these subjects in separate buildings with separate syllabi. Pharmacy students should learn data science. Engineering students working on AI should spend time in hospitals and laboratories. The National Education Policy 2020 encourages this kind of multidisciplinary study, and universities that bring medicine, pharmacy and computing onto one campus are well placed to lead.
Finally, someone must be answerable when an algorithm makes a mistake. If an AI system misses a tumour or recommends the wrong treatment, the patient deserves to know who is responsible. Clear rules on liability will build public trust faster than any advertisement.
What the reader can expect
For ordinary citizens, AI in the life sciences will rarely look dramatic. It will appear as a quicker X-ray report at the district hospital, a crop variety that survives a dry spell, or a medicine for a neglected disease that reaches the market a few years sooner. It is fair to ask your doctor whether an AI tool was used in your care and how its result was checked. Asking such questions is a healthy habit, and good doctors will welcome it.
Think back to the woman coughing into the health worker’s phone. The app’s verdict is worth something when it leads to a proper test, a correct diagnosis, free medicines and a health worker who visits until she is cured. The algorithm opens the door. Nurses, doctors, laboratory staff, scientists and sound public policy have to carry her through it.
The author is a professor of computer science and engineering at Jamia Hamdard, New Delhi, and served as its vice chancellor from 2021 to 2026; Views presented are personal.