AI Transforms Drug Discovery Amid Legal Uncertainty Over Patent Rights and Profit Sharing

Artificial intelligence is set to transform drug discovery, clinical trials and regulatory processes, marking a new scientific revolution in biotechnology, but legal uncertainty surrounding patent rights and profit distribution is growing as the technology reshapes pharmaceutical development.

According to a report in the international academic journal Science on the 17th (local time), intellectual property experts have pointed out that issues of ownership and profit sharing in AI-driven drug development remain in a legal vacuum.

AI can accelerate the drug development process by rapidly screening millions of molecules and analyzing potential problems in the clinical process early, thereby reducing the time required for clinical trials. Although no drug discovered and designed by AI has yet reached the market, two new drug candidates developed using AI have confirmed their safety in Phase 2 clinical trials: Insilico Medicine's pulmonary fibrosis treatment Lentosertib and Recursion's cerebral cavernous malformation treatment REC-994. Currently, it is known that over 90% of new drug candidates fail in clinical trials, meaning that roughly only one out of ten candidates successfully reaches commercialization and is prescribed to patients.

The problem is who holds the patent rights when AI plays a core role in the drug development process. In the 2022 Thaler v. Vidal case, a U.S. federal appeals court ruled that an AI cannot be registered as an inventor on a patent. As an appellate court decision, it can only be overturned by the U.S. Supreme Court. The likelihood of the Supreme Court overturning the decision is low.

The Thaler v. Vidal ruling did not address practical situations involving collaboration between AI and humans. In November of last year, the U.S. Patent and Trademark Office published revised guidelines in the Federal Register regarding inventorship for AI-assisted inventions, specifying that AI should be treated as a tool, similar to laboratory equipment or software.

Applying these legal standards to actual drug development complicates matters. Whether someone is recognized as an inventor can vary depending on the type of molecule. Even if an AI proposes the digital form of a molecule, the synthetic chemist who devises the actual synthesis route is likely to be recognized as the inventor. It is also pointed out that for macromolecules like proteins or nucleic acids, proving human contribution is more difficult because the methods for manufacturing them from AI-discovered amino acid sequences are already well-established.

The way AI usage is documented is also a potential risk factor. While humans are poor at recording their thought processes, the outputs from an AI can be easily saved, potentially exaggerating AI's contribution. During patent litigation, when reviewing lab notebooks and public data, a claim could be made that the AI was the inventor and the human did not contribute. This suggests that the level of AI utilization must be accurately documented.

It is unlikely that AI development companies will claim rights to inventions made using their AI. Currently, most AI model providers assign the rights to the outputs to the user in their service agreements. In high-value fields, separate contract negotiations may occur over arrangements like sharing profits based on the drug's sales performance.

The balance of power between parties in contract negotiations is also a variable. When a large AI company negotiates with a small pharmaceutical company, the AI company may propose more favorable terms. When a large pharmaceutical company negotiates with a small AI company, the pharmaceutical company is likely to take the lead. The profit-sharing structure can vary depending on the size and negotiating power of both sides.

Speaking at the BioAsia 2026 summit, a chief scientific officer and senior vice-president of global research said AI is fundamentally changing how science is conducted across research and development. "Science changes when a fundamental technology disrupts the way we understand and work. AI is now driving such a model revolution in our industry," the executive said.

AI is being used from the earliest stages of molecule discovery to clinical development. Companies have invested in computational infrastructure, including advanced supercomputing systems, to analyse laboratory data and biological images at scale. AI tools help identify patterns, predict outcomes and accelerate decision-making in drug development and clinical trials.

Highlighting challenges in understanding complex diseases, the executive said human genetics plays a key role. Large-scale genomic datasets covering hundreds of thousands of individuals are now being analysed using machine learning models to identify disease-causing mutations. "AI enables us to test millions of mutations computationally, something that was previously not feasible," the executive said.

AI models have been trained to distinguish between harmful mutations and incidental genetic variations in cancer research. Deep learning is also being applied to design DNA sequences that can control gene expression in specific cell types. AI-designed biological sequences have in some cases outperformed natural sequences in laboratory experiments. This has been described as a form of information compression, where complex biological rules are distilled into efficient, functional designs.

In manufacturing, AI is helping optimise production processes and improve efficiency by modelling multiple variables simultaneously. The integration of AI with automation, large-scale biological data and collaborative research is accelerating the pace of therapeutic innovation. "There has never been a better time for drug discovery. AI is reshaping the future of medicine," the executive said. The executive also spoke about the ambition of zero-shot antibody design, where computational models can generate therapeutic antibody candidates directly from a target sequence without traditional experimental screening.

Researchers are harnessing AI to fast-track discoveries, offering fresh insights into life at the molecular level and new strategies against disease. A team has developed D-I-TASSER, a new software tool that predicts the 3D shapes of complex proteins more accurately, supporting faster drug discovery, improved disease research and more precise design of targeted therapies.

"For most proteins, we still do not know their 3D structures, and that remains a major blind spot in biology," a professor said. "A protein's shape determines what it does in the body, but many large, multi-domain proteins are too complex for existing tools to model reliably." The human body contains about 20,000 different proteins, many of which consist of several connected parts that move and interact with each other. This complexity makes accurate computer modeling difficult and slows progress in understanding disease mechanisms and developing new medicines.

To address this challenge, the team developed D-I-TASSER by combining AI with physics-based simulations. The system breaks a complex protein into smaller sections, predicts the shape of each section first, and then uses physical modeling to assemble them into a complete three-dimensional structure, allowing more precise reconstruction of how the protein folds and fits together. In tests, D-I-TASSER predicted complex protein structures about 13% more accurately than existing state-of-the-art methods. The researchers were also able to generate reliable structural models for most proteins in the human body, including many that were previously difficult to analyze.

"When we can see a protein's structure more clearly, we can better understand what goes wrong in disease and how potential drugs might interact with it," the professor added. Building on D-I-TASSER, the team is extending the framework to RNA structure prediction and to modeling protein–protein interactions, with a particular focus on antibody–antigen complexes.

As biotech and biopharmaceutical companies fully embrace the AI era, this technology is demonstrating its profound value in the medical field. Google DeepMind, through its AlphaFold project, solved the protein folding problem and won a Nobel Prize, clearing a critical obstacle on the path to AI-driven drug discovery. As AI models continue to iterate and upgrade, customization costs keep decreasing, and deployment barriers for enterprises gradually lower, the tipping point for technological breakthroughs is approaching year by year.

Recursion Pharmaceuticals, focused on AI-driven drug development, has attracted investment from prominent technology investors. Although the company's stock price has dropped by 90%, currently giving it a market cap of $1.9 billion, an analyst upgraded its rating last December, primarily optimistic about its candidate drug REC-4881 for treating familial adenomatous polyposis. The analyst pointed out that several candidates in its AI-powered R&D pipeline have been validated by clinical results and pharmaceutical partnerships.

A major pharmaceutical company recently established a $100 million collaboration with NVIDIA to build a joint innovation lab. Empowered by top-tier computing capabilities enabled by NVIDIA's Vera Rubin chips and the BioNeMo platform, this biopharma giant is accelerating its release of AI potential. Currently valued at under $1 trillion, the company is showing attractive valuation appeal.

While AI can significantly accelerate the pace of drug development, progress may be delayed if patent ownership and contract structures are not sorted out. Experts emphasize that "these legal issues must be resolved quickly to accelerate development and ultimately benefit patients."