Rentosertib, the first drug conceived entirely by AI, is not only moving into the final phase of clinical trials, but is rewriting the rules of the game.
In less than two years, a treatment for idiopathic pulmonary fibrosis reached the stage where humans become testing ground.
“Biology is becoming computational”
By: Gabriel E. Levy B.
In a Phase 2 clinical trial conducted in China and published by Nature Medicine in June 2025, Rentosertib showed promising results in patients with idiopathic pulmonary fibrosis, a devastating disease that kills about 40,000 people a year in the United States alone.
The data were clear: those who received the highest dose significantly improved their lung capacity compared to the decrease in the placebo group. So far, one more clinical success.
The disruptive, however, is not so much in the result as in the form.
Insilico Medicine, the company behind the discovery, managed to accelerate the process from design to human trial in just 18 months.
A feat if you consider that the average to reach that stage exceeds four years. It all started with a library of 78 molecules and an AI platform capable of identifying patterns in biochemical chaos.
According to Insilico’s founder, Alex Zhavoronkov, the use of AI at this stage is not an auxiliary tool, but the new engine of biomedical research.
The philosopher and technologist Luciano Floridi warned a few years ago that “biology is becoming computational” and that, therefore, “medicine will cease to be an artisanal science to become a science of prediction based on data” (The Logic of Information, 2019).
Rentosertib is not just a molecule, it is the first ambassador of this new paradigm.
“This is the year things started to work”
The announcement that Rentosertib will enter Phase 3 in the next 18 months coincides with a tide of investor enthusiasm.
During the first three quarters of 2025, venture capital injected more than $2.7 billion into AI-based drug discovery startups, according to data from PitchBook.
This is not an isolated fact. Projected global investment for this sector will reach $2.51 billion by 2026 and could grow to $16.49 billion by 2034, with a compound rate of 27%.
In this fast-paced ecosystem, money flows to those companies that promise not only to shorten R+D timelines, but also to reduce failure rates in advanced clinical phases.
Eli Lilly, for example, teamed up with Nvidia to build what will be, in January 2026, the most powerful supercomputer ever installed in a pharmaceutical company.
It will not be a factory of pills, but of data: millions of molecules simulated, evaluated and discarded in minutes, something that used to take years.
Josh Meier, co-founder of OpenAI-backed company Chai Discovery, summed up the new zeitgeist by saying, “This is the year things started to work.”
With just two rounds of funding, the startup reached a valuation of $1.3 billion. And in the world of biotechnology, where uncertainty used to be the norm, the word “work” is invaluable.
For sociologist and science theorist Bruno Latour, laboratories are not neutral spaces, but “networks of inscription where scientific knowledge is transformed into stable facts.” With AI, those networks are no longer physical.
They are distributed in clusters of GPUs, and the “stable facts” are now viable molecules discovered by algorithms that do not sleep or doubt.
The algorithm as a biochemist
AI doesn’t just look for combinations of compounds. Its true promise is to understand the complex patterns that govern the human body on a scale impossible for the human brain.
DeepMind, for example, dazzled the world in 2020 with AlphaFold, a system that predicted the structure of more than 200 million proteins. Now, that knowledge is not only stored, it is applied.
In the case of Insilico, its platform integrates data from medical images, genetic sequences, previous clinical studies and molecular simulations.
What were once separate tasks between teams of chemists, biologists, and clinicians are now intersected and integrated in real time thanks to the power of computing.
This change is not only methodological, it is ontological.
Traditional medicine was a science of trial and error, where discovery was as random as a needle in a haystack.
With AI, the haystack becomes a database, and the needle becomes an algorithmic prediction.
However, not everything is certain. Rentosertib’s own success in the early stages does not guarantee that it will make it past Phase 3, where many promising candidates have historically failed.
In addition, there is the ethical debate: can algorithms truly understand human complexity or only replicate it? Are we creating knowledge or simply optimizing patterns?
Epistemologist Sabina Leonelli warns about the risks of knowledge automation in biology. In his text Data-Centric Biology (2016), he points out that “reliance on large volumes of data cannot replace the need for contextual interpretation”.
In other words, the algorithm can find molecules, but medicine is still a profoundly human science.
Candidates who are already in the race
The case of Rentosertib is not an isolated exception.
Retro Biosciences, fueled by a $180 million personal investment by Sam Altman, has already begun clinical trials with an experimental Alzheimer’s drug.
In parallel, companies such as Valo Health, Recursion Pharmaceuticals, and Atomwise are working on treatments for diseases ranging from cancer to rare diseases, all with an AI-based approach.
In December 2025, Chai Discovery raised $130 million in a Series B round and claims to have generated molecules with “unprecedented” success rates.
Beyond the optimistic headlines, the important thing is that there are already molecules that have entered clinical phases, that is, they are being tested in humans. That is the critical threshold between promise and reality.
Insilico, for his part, did not stop at Rentosertib. In January 2026, it announced a partnership with Hisun Pharmaceutical to nominate a new preclinical candidate in just eight months, a speed that would have been unthinkable in the pharmaceutical industry a decade ago.
Even large companies such as Pfizer and Novartis have begun to publish studies on their own AI-assisted discovery platforms, although they have not yet presented clinical results as advanced as those of Insilico.
In conclusion
Rentosertib is not just a molecule on its way to becoming a drug, it is the emblem of a new era in biomedicine.
Artificial intelligence is no longer a promise of the future, but an active tool in drug discovery.
Its effectiveness, still under evaluation, has caused a radical transformation in time, costs and even in the way of understanding research. The question is no longer whether AI can create medicines, but how it will reshape the way we understand disease, the body, and science itself.
References:
- Floridi, Luciano. The Logic of Information: A Theory of Philosophy as Conceptual Design. Oxford University Press, 2019.
- Latour, Bruno. Life in the laboratory. Alianza Editorial, 1995.
- Leonelli, Sabina. Data-Centric Biology: A Philosophical Study. University of Chicago Press, 2016.
- Nature Medicine, June 2025. Rentosertib Phase 2 results.
- PitchBook Data Inc., report on investing in biotechnology with AI, 2025.
- Precedence Research, Global AI in Pharma Market Size and Forecast 2024-2034, published in 2025.
