In all the worthy discussions around the promise and peril of AI, we may be overlooking one of its most powerful use cases: solving urgent global health crises. Few problems illustrate this better than antibiotic resistance.
Antibiotics underpin modern medicine, enabling procedures like C-sections and organ transplants and ensuring that patients can safely receive treatments such as chemotherapy. But the bacteria they target are constantly evolving. Over time, many have developed resistance to the drugs we rely on—turning once-routine infections into life-threatening conditions.
The scale of the problem is staggering. A landmark global analysis published in The Lancet estimates that antibiotic-resistant infections—known as superbugs—could directly cause more than 39 million deaths between now and 2050, with resistant bacteria contributing to more than 8 million deaths per year by mid-century if current trends continue. In 2019 alone, antibiotic resistance was responsible for 1.2 million deaths globally, exceeding the toll of AIDS-related illnesses and malaria that year.
At the same time, the pipeline for new antibiotics has been shrinking for decades. Traditional drug discovery is slow, expensive, and notoriously inefficient. Scientists must test thousands, or even millions, of chemical compounds to identify just a few viable candidates, according to our internal research.
A PROBLEM FOR AI
This is exactly the kind of problem AI is built to tackle.
Antibiotic discovery represents an ideal use case for artificial intelligence that can serve as a paradigm for AI drug discovery more broadly. Instead of testing molecules manually, AI models can analyze vast chemical libraries to predict and even design compounds which are most likely to kill bacteria, dramatically narrowing the field before a single experiment begins.
The result is faster and better research.
Across the emerging field of AI-native drug discovery, there is growing consensus that machine learning can reduce the timeline of the early discovery phase—covering hit identification, hit-to-lead optimization, and lead optimization—by 50% to 75%, in our experience. That means moving from a promising molecule to a preclinical drug candidate in a fraction of the traditional pace.
But speed is only part of the story.
AI dramatically expands the chemical universe scientists can explore. This is particularly important in antibiotic discovery, where many existing drug scaffolds are already vulnerable to well-understood bacterial resistance mechanisms. To stay ahead, researchers must identify entirely new scaffolds and mechanisms of action.
INCREASE SHOTS ON GOAL
Traditional discovery methods limit researchers to relatively small collections of molecules that can realistically be synthesized and screened in the lab. AI models, by contrast, can explore tens to hundreds of millions of potential compounds in silico—computer modeling. It can then prioritize the most promising candidates for synthesis and experimental testing, helping surface entirely new chemical structures that researchers might not have otherwise considered.
In other words, AI increases the number and quality of “shots on goal.”
Crucially, this technology exists to amplify human intelligence—learning from and augmenting the insight and judgment of scientists.
At organizations like ours, Phare Bio, and across the broader biotech ecosystem, AI is being used as a collaborative tool. Machine learning models generate hypotheses, prioritize molecules, and analyze patterns in biological data. Researchers then validate those predictions in the lab, refine the models, and guide the next iteration of discovery.
This partnership between human and machine intelligence is already reshaping multiple areas of drug development.
Some companies focus on small molecule chemistry, using AI to reduce the number of compounds that must be synthesized and tested. Others are designing entirely new biologic medicines, such as antibodies, where machine learning can accelerate the traditionally slow process of antibody discovery. Still others apply AI to simulate protein dynamics, helping researchers understand how molecules interact with dynamic biological targets.
These approaches may differ technically, but they share a common goal: finding better drugs, faster.
LOWER THE SCIENTIFIC DISCOVERY BARRIER
Perhaps most importantly, AI is lowering the barriers to entry for scientific discovery.
Historically, antibiotic research required enormous infrastructure: large pharmaceutical companies, massive screening libraries, and expensive laboratory pipelines. Today, powerful AI models and open datasets allow smaller teams, such as academic labs, nonprofits, and startups, to compete in the race to find new antibiotics.
That democratization matters. Antibiotic resistance is a global problem that requires a global response.
For all the hope surrounding artificial intelligence, its greatest impact may ultimately come from helping humanity solve problems that once seemed intractable.
Antibiotic resistance is one of the most serious biological threats we face. But it is also a challenge uniquely suited to the strengths of AI: pattern recognition, massive-scale exploration, and rapid iteration.
If we continue building smarter models, pairing them with human expertise, and applying them to the urgent challenges of global health, AI could help unlock an entirely new generation of antibiotics.
Akhila Kosaraju, MD, is CEO and cofounder of Phare Bio.