AI is discovering new drug candidates in record time, but the legal system hasn't caught up with the technology. While companies like Insilico Medicine are bringing drug candidates from conception to human trials in under 30 months (compared to the traditional four to six years), a fundamental question remains unanswered: who actually owns these AI-generated inventions? Patent offices in Europe, Germany, and the United States have all ruled that AI cannot be listed as an inventor, but the practical consequences of that decision are creating a legal minefield for pharmaceutical and biotech companies racing to use these tools.

What Do Patent Offices Actually Say About AI Inventors?

The legal position is surprisingly clear, even if it creates headaches for companies. The European Patent Office (EPO) established in 2020 that only natural persons can be inventors, and AI can at most be mentioned as a tool. Germany's Federal Court of Justice confirmed this in June 2024, and the United States Patent and Trademark Office (USPTO) followed suit with updated guidelines in 2024. The ruling sounds like a technicality, but it has major real-world implications.

For a patent to hold up in court, companies must now prove that a human made a substantial inventive contribution to the discovery. This means documenting exactly where human judgment came into play, whether in target selection, interpreting results, or refining the candidate. Simply typing a prompt into an AI tool is not enough. The challenge is that the legal system has not yet defined precisely how much human contribution qualifies as "substantial," leaving gray areas that competitors and licensing partners can exploit.

How Should Companies Protect AI-Generated Discoveries?

The stakes are high enough that companies need to act now. Internal documentation has become mandatory, not optional. Development processes must be recorded in detail so that the human contribution to each invention can be traced if a dispute arises. But documentation is only part of the puzzle. Contracts matter just as much, and many companies are overlooking the most critical one: the agreement with the AI provider itself.

• Document Development Processes: Internal workflows should be designed to record the human contribution at every AI-assisted step, creating a traceable record for patent applications and intellectual property defense in case of dispute.
• Review Contracts with AI Providers: Before signing up for an AI platform, companies must clarify who owns the generated results, models, and data. This contract often determines the economic allocation of results and is frequently the least carefully negotiated agreement in the entire process.
• Update Existing License Agreements: Companies that introduced AI-assisted research after signing earlier partnership or licensing deals should examine whether existing IP definitions still reflect the new technical reality, especially in co-development scenarios.

Classic license agreements were designed for a world where humans made all the inventions. They typically define licensed intellectual property through patent lists or descriptions of know-how. But AI-generated IP often includes the underlying training data, models, and algorithms. If a contract does not explicitly include or exclude these elements, it opens the door to conflict. When two companies jointly conduct AI-assisted research, the question of who owns the result becomes even murkier, especially if one party provides the AI platform and the other contributes biological or medical expertise.

Can Quantum Computing Make AI Drug Discovery Even Better?

While companies grapple with legal ownership, researchers are already working on the next frontier: combining AI with quantum computing to make drug discovery even more powerful. A new peer-reviewed study published in Scientific Reports shows that annealing quantum computers can meaningfully improve how AI designs new drug molecules. The research was conducted by D-Wave and Shionogi, a major Japanese pharmaceutical company.

The problem that generative AI faces in drug discovery is deceptively simple to state but hard to solve. The space of chemically plausible, drug-sized molecules is estimated at more than 10 to the 60th power, while humans can currently synthesize and test only about 10 to the 10th power. Generative AI models trained on the tiny slice of molecules that humans have already made tend to overfit, producing outputs that are either invalid chemistry or poor drug candidates. This is called the "drug-likeness problem," and it remains one of the main reasons AI-generated molecules still need heavy filtering before any can be synthesized.

The D-Wave and Shionogi team built a generative model that used a D-Wave annealing quantum computer as part of its sampling process. They compared it head-to-head with an identical model running on classical hardware only. The quantum-assisted version produced chemically valid molecules 97 percent of the time, compared with 73 percent for the same architecture on classical hardware and 54 percent for a previously published benchmark model.

On drug-likeness, measured by a standard metric called the Quantitative Estimate of Drug-Likeness (QED) score, the results were even more striking. The quantum-assisted model generated drug-like molecules at a rate of 66.79 percent, compared with 43.15 percent for the best classical-only model and 31.61 percent for the original training data. In other words, the quantum model generated drug-like molecules at more than double the rate present in the training data.

The implications reach far beyond drug discovery. Based on these results, annealing quantum computing could act as a general-purpose sampling engine inside generative AI, drawing samples more effectively from the "in-between" spaces that a model has not directly seen in training. This is exactly the capability that limits generative AI across many domains, not just molecular design. D-Wave's annealing quantum computers are already being used in customer applications across manufacturing, telecommunications, retail, logistics, and defense, addressing real-world optimization problems today.

What Happens Next for AI Drug Discovery?

The legal landscape surrounding AI-generated intellectual property will evolve dynamically over the coming years, but companies that act now have a clear advantage over those that wait. The technology is already here. Recursion Pharmaceuticals, which merged with Exscientia in late 2024, is advancing multiple AI-powered drug candidates in oncology and rare diseases. Insilico Medicine is now in Phase II trials with its AI-discovered candidate for idiopathic pulmonary fibrosis.

The convergence of faster legal clarity, better contract structures, and improved AI and quantum computing capabilities suggests that AI-assisted drug discovery will only accelerate. But the companies that succeed will be those that treat intellectual property strategy and contract negotiation with the same rigor they apply to the science itself. The invention may be AI-assisted, but the legal protection depends entirely on human foresight.