How artificial intelligence is changing medicine
Despite the hype, speculation, and alarming forecasts, few experts doubt that artificial intelligence will indeed change the world. However, who will benefit from these changes and what price will have to be paid remain open questions.
History shows that technological breakthroughs often bring crises along with opportunities, forcing society to seek a new balance. Yet, there is one field where the benefits of technological progress have been almost indisputable for decades: medicine.
ForkLog explored how the application of artificial intelligence is already accelerating the development of new drugs, optimizing laboratory processes, enhancing diagnostic accuracy, and changing treatment approaches for various diseases.
Drug Development
Most medications work by interacting with receptor proteins—molecular structures that regulate cell function and participate in nearly all bodily processes.
AI systems can analyze the structure of receptor proteins and predict which compounds will interact with them most effectively and with minimal side effects. As a result, tasks that previously required many years of laboratory research are increasingly being solved in just months.
According to estimates from the World Health Organization (WHO), most new pharmaceutical drugs will be developed using AI in the coming years.
AlphaFold and Isomorphic Labs
In 2024, the Nobel Prize in Chemistry was awarded to David Baker, Demis Hassabis, and John Jumper. The latter two work at Google DeepMind and were recognized for developing methods to predict protein structures, including AlphaFold, which is based on machine learning.
In 2018, AlphaFold took first place in the "competition" for molecular prediction known as the Critical Assessment of Structure Prediction (CASP), demonstrating effectiveness in the most challenging categories. Two years later, a new version—AlphaFold 2—won the next CASP.
In 2021, Google DeepMind released the AlphaFold2 code and a database of predicted protein structures to the public. Around the same time, Hassabis founded Isomorphic Labs, a subsidiary of Alphabet focused on developing AI for drug discovery.
In 2024, Isomorphic Labs secured partnerships with Eli Lilly and Novartis, which included funding for the company’s AI research of up to $1.7 billion and $1.2 billion, respectively. In 2026, Isomorphic Labs also announced a partnership with Johnson & Johnson.
In February 2026, Isomorphic Labs unveiled a universal drug development environment called the Drug Design Engine (IsoDDE), built on AlphaFold technologies.
Currently, Isomorphic Labs is working on solutions in oncology and immunology. Despite the acceleration of development through AI, projects remain in the preclinical research stage. The company expects to begin human trials in the coming years.
Exscientia and Recursion Pharmaceuticals
Founded in 2012, Exscientia became one of the first companies to systematically apply machine learning to drug development.
In 2020, the drug DSP-1181 for treating OCD became the first AI-created product to enter clinical trials. The development was conducted in collaboration with the Japanese pharmaceutical company Sumitomo Dainippon Pharma, which handled synthesis and laboratory testing based on Exscientia's theoretical results.
By 2023, the company had prepared eight candidate molecules developed "significantly faster" than the industry average.
In 2024, Recursion Pharmaceuticals acquired Exscientia in a $688 million deal, closing some research programs.
By that time, several drugs had reached the second stage of clinical trials—testing efficacy and side effects on groups of 100–300 patients.
The merger with Recursion Pharmaceuticals allowed the use of Exscientia's AI systems in combination with an automated laboratory complex for testing. Additionally, Recursion built its own AI supercomputer, BioHive-2, using NVIDIA H100 to train specialized models.
The company also participated in developing an open generative model, Boltz-2, designed to predict the three-dimensional structure of proteins.
By 2025, Recursion Pharmaceuticals focused its efforts on four oncology programs and two related to rare diseases. Several drugs are already in transition between the first and second phases of trials:
- REC-4881 for treating familial adenomatous polyposis, a condition that increases the risk of colorectal cancer;
- REC-617 for treating malignant ovarian tumors;
- REC-1245 for combating lymphoma and other forms of malignancies.
The drug REC-3565, intended for treating chronic lymphocytic leukemia, is undergoing the first stage of clinical trials.
Insilico Medicine
Founded in 2014, Insilico Medicine is another significant player in AI-driven drug development.
In 2017, Insilico Medicine was listed among the top five projects for social impact by Nvidia.
The company utilizes AI at all stages of the development cycle:
- The PandaOmics system identifies biological "targets"—molecules that need to be "turned off" or regulated in therapy;
- Chemistry42 provides generative design of suitable compounds;
- InClinico optimizes predictions for clinical trials.
One of Insilico Medicine's early AI achievements is the drug Rentosertib (ISM001-055), related to fibrosis treatment. The development took 18 months from target discovery by the AI system to obtaining a candidate molecule. As of 2025, Rentosertib is undergoing the second phase of clinical trials.
Additionally, in 2024, the AI-developed immunomodulatory drug ISM3312 for COVID-19 and other viral infections passed the first phase of trials. ISM3091, related to cancer therapy, has been approved for patient testing.
Diagnostics and Research
Experts estimate that about 90% of all medical information is represented by images such as X-rays and tomograms. This data is critical for diagnostics, but analyzing it is a labor-intensive and non-trivial task.
Machine learning methods, especially convolutional neural networks, are well-suited for recognizing complex visual patterns. Similar to human vision, these systems can distinguish contrasting edges, shapes, and textures in images. This allows for the confident identification of tumors, hemorrhages, and other anomalies.
High-quality data sets—collections of documented images with expert annotations—are available for training AI models.
In 2024, researchers from Harvard Medical School introduced an AI model named Chief, capable of detecting several forms of cancer. According to the developers, the solution accurately identified signs of disease in digital images 94% of the time.
In 2025, the U.S. Food and Drug Administration (FDA) designated the Damo Panda model from Damo Academy—Alibaba's research division—as a "breakthrough device." According to the developers, the system can recognize signs of pancreatic cancer on tomograms even before symptoms appear, which is particularly important for this type of disease.
In 2026, a significant breakthrough in AI diagnostics was the REDMOD system, developed by the American non-profit Mayo Clinic.
This model, also designed to detect pancreatic cancer, outperformed specialists in diagnosing the disease at early stages. According to researchers, the system identified pathological changes on tomograms an average of 475 days before diagnosis.
Google Initiatives
Google is a key provider of AI for medical diagnostics and research.
The company offers a range of open models for analyzing medical texts, images, and audio through MedGemma, based on Gemma 3.
Through Health AI Developer Foundations, developers have access to open weight sets and AI tools.
Google collaborates with various clinics and research organizations, focusing on developing foundational technologies.
In 2019, the company presented a model for detecting and predicting lung cancer. The model performed on par with or better than a group of six certified radiologists.
In 2020, in collaboration with Northwestern Medicine, researchers demonstrated a system for analyzing mammograms capable of detecting cancer at the level of a specialist.
In 2024, Google Cloud and the German pharmaceutical company Bayer announced the launch of a platform for screening X-rays. The system analyzes the history of images and data from medical records, forming hypotheses about possible pathologies.
NVIDIA and GE HealthCare Radiologist Robots
Tech giant Nvidia and American medtech company GE HealthCare, which produces radiographic equipment, are developing their own AI system for autonomous imaging.
Unlike models that analyze already prepared images, this solution aims to reduce the routine burden on specialists and standardize diagnostics.
Initially, the system will work with X-rays and ultrasound images.
GE HealthCare also plans to utilize NVIDIA Isaac for Healthcare—a platform for developing autonomous medical systems, including surgical robots.
PathAI Diagnostic Platform
Founded in 2016, PathAI developed a "digital pathology platform" called AISight Dx, designed for primary diagnostics in clinical settings.
The system provides an environment for working with medical images, allowing the integration of third-party algorithms for data analysis.
It claims support for a set of CE-IVD certified AI-based solutions, including "plugins" for oncological diagnostics:
- DeepDx Prostate automatically highlights tissues in images and identifies potentially important areas for diagnosis;
- Histotype Px Colorectal builds predictions for disease progression based on images, assesses the appropriateness of chemotherapy, and offers therapeutic recommendations;
- Visiopharm identifies and counts biomarkers for various forms of cancer.
The platform includes its own functions for automatic image analysis, assisting in diagnosis formulation and report writing, but these are currently intended "exclusively for research purposes" and are not approved for clinical use.
AISight Dx also offers built-in AI tools:
- ArtifactDetect—for finding scanning artifacts and other errors in images;
- Case Priority—for prioritizing clinical cases based on tissue analysis;
- AIM-Tumor Cellularity—for assessing the cellular composition of tumors.
In 2022, the solution received FDA 510(k) clearance and the European CE mark, indicating the product's safety for consumers and the environment.
In 2025, PathAI announced a partnership with the Moffitt Cancer Center in Florida, USA, to implement AISight Dx in diagnostic processes. In 2026, the company entered into a similar agreement with the University Hospital Zurich.
In May 2026, Swiss pharmaceutical company Roche announced its acquisition of PathAI in a deal worth over $750 million.
Challenges and Limitations
Like other industries, the use of AI in medicine exacerbates systemic issues and creates new ones.
AI assistants, especially those based on LLMs, are not immune to hallucinations.
In a research paper from Google about the Med-Gemini model, an error was found: the model "invented" a non-existent area of the brain called the basal nuclei.
This hallucination arose from two real anatomical names: basal ganglia and basilar artery. Developers attributed it to a typo, but several experts called the incident a concerning example of the risks of implementing AI assistants in medicine.
Researchers from Stanford University found that AI models could convincingly diagnose diseases from medical images without access to the images themselves.
One analyzed system "blindly" achieved high results in a radiology test. Models like GPT-5, Gemini 3 Pro, and Claude Opus 4.5 "confidently described visual details" on non-existent images.
According to a study published in June of the same year in a medical context, 7.1% of GPT-4's responses to patient questions were incorrect and could have caused significant harm. In one case out of 156, the error posed a risk to life.
By 2025, tools for automatically documenting dialogue results with patients made errors in 70% of clinical notes. Models added false facts to the conversation transcript, omitted key points, and confused concepts.
In addition to LLMs fabricating organs, they exhibit a lack of transparency in their logic, making it difficult for humans to analyze how certain conclusions were reached.
A lack of representativeness in datasets can create biases and lead to false patterns in the models trained on them.
Moreover, typical issues with AI assistants, such as cognitive dependency of users and data privacy, are only exacerbated in the healthcare context.
WHO experts consider the use of artificial intelligence in medicine to be a high-risk area.
Under the European AI Act, starting in August 2026, AI systems in this category will be required to meet a series of special requirements related to risk management, reporting, and human oversight.
Despite the challenges and potential risks of implementation, WHO views the prospects of artificial intelligence in medicine positively, provided there are appropriate regulations and oversight from government agencies.
The FDA in the U.S. is also optimistic about the prospects of medical AI, although they acknowledge that existing regulations are outdated. Formally, such systems in the U.S. are classified as Software as a Medical Device.
In 2025, the FDA published a set of recommendations regarding the lifecycle of AI products, risk management, and marketing.