One promise, perhaps most important, has been made by some artificial intelligence (AI) lab leaders that their technology will contribute Cure for many diseases. “I think AI can cure the biggest diseases in the next five to ten years.” These words were written by Dario Amodei, CEO of Anthropic on September 13.
Demis Hassabis, CEO of Google DeepMind, defended this same premise in April 2025: “We can cure all diseases with AI. Maybe in the next decade. I don’t see why.” And Sam Altman, the head of OpenAI, also recently expressed his wish: “I don’t want to stop technological progress. I want a cure for that cancer, for Alzheimer’s.”
These statements are very encouraging, but for now they are only promises. Despite how fast AI is advancing, the expectations of technology company leaders are still far from the contribution it is making to the development of new drugs. This technology is proving to be a very valuable tool in this field, but it is a long way from putting effective and safe medicine in our hands. And even further away from curing cancer or Alzheimer’s.
Vertex test
Jeff Layden, president of the pharmaceutical company Vertex, explained that his laboratory has turned to about 50 AI companies to analyze a database of about 10,000 drug compounds. The results obtained were very disappointing: none of the AI models could produce an effective drug. This is not to say that this technology has nothing to say when it comes to developing new drugs, but it does put an unpleasant reality on the table: AI today cannot discover new drugs on its own.
The human body is extraordinarily complex because thousands of biological processes intervene in it.
Furthermore, Liden makes an important reflection: AI can act as a co-pilot for researchers today. You can help them, but Can’t replace them. To develop a new drug, it is not enough to find a molecule capable of interacting with a specific protein. The human body is extraordinarily complex because thousands of biological processes intervene in it. And many of them have very deep dependencies that are hard to identify and understand.
AI models have the ability, above all, to recognize patterns, find potential therapeutic targets, predict protein structures or suggest promising molecules. But so far none of them have been able to predict with sufficient accuracy how a compound will behave when exposed to a human organism. And this knowledge is essential.
Preclinical trials and clinical research phases can last several years
Moreover, there is another concept that we should not ignore: the drug development process is more complicated when it is necessary to predict side effects and interactions with biological processes. In any case, let’s imagine that an AI manages to create a very promising molecule that, on paper, does not have harmful side effects or trigger potentially dangerous interactions with some biological process. Still, there will be a loose end: Clinical trials.
This phase of new drug development is crucial and cannot be avoided no matter how well AI models predict that a drug works correctly and is not dangerous. Preclinical trials and clinical research phases can last several years, so accelerating the discovery of drug candidates does not necessarily mean increasing the proportion of drugs reaching the market. The pharmaceutical industry is using AI, yes, but its expectations are far more conservative than those of AI lab leaders.
Picture | Artem Podrezh
More information The Wall Street Journal
In Xataka | A section of the mathematical community is up in arms against OpenAI
