When AI Discovers the Molecule, Who Owns the Invention?
- Rajashree Rajadhyax
- Aug 24
- 3 min read
A fascinating story originally reported by MIT Technology Review recently caught my eye. It highlights a compelling friction point where corporate messaging, medical innovation, and strict intellectual property rules intersect.
When biotech company Insilico Medicine announced that it had developed a promising new drug candidate for idiopathic pulmonary fibrosis, a serious lung condition, it made headline news. In press releases and presentations, the team shared that the molecule had been generated using its AI platform ‘Pharma.AI’. It was widely covered as a major milestone for computational drug discovery.
In press releases, presentations, and interviews, the team proudly shared that the drug had been "discovered by" generative AI. It was a major milestone for the technology, showing just how fast computational models can process complex biological data and suggest potential treatments.
However, when it came time to file the official patent for the exact same molecule, something interesting happened. The AI wasn't listed on the document at all. Instead, the legal application named five human scientists as the drug's inventors.
So why the sudden change in narrative? Why the structural difference between the public announcement and the legal filing? Is it simply a matter of marketing vs. legal necessity, or does it reflect a broader paradox?
The Law: Humans Only
To understand what happened, we have to look at how patent laws work around the world. Patent offices like the United States Patent and Trademark Office (USPTO), the European Patent Office, and various legal authorities across Asia and the UK operate under a strict rule: only a "natural person", a human being can be legally named as an inventor.
This rule isn't just an overlooked technicality; it has been repeatedly tested in court. In recent years, researchers tried to register patents where an AI system called DABUS was listed as the sole inventor. Patent offices and judges across multiple countries rejected the applications, ruling that current patent statutes were written explicitly to protect and reward human ingenuity.
If a company tries to list an AI software program as an inventor on a patent application today, the patent office will simply reject it outright.
Where Do Humans Fit In?
Does this mean AI-generated drugs cannot be patented at all? Not quite. Recent guidance from patent regulators clarifies that using AI as a tool does not disqualify an invention from receiving a patent. However, the human applicants must show that they provided a "significant contribution" to the final creation.
In the case of AI drug discovery, that human contribution usually includes several critical steps:
Framing the Problem: Human scientists define the specific biological targets, curate the training data, and set the precise parameters for the algorithms.
Lab Testing and Validation: While an AI model might propose thousands of theoretical chemical molecules, human researchers must evaluate those options. They decide which specific compounds are worth creating in a real laboratory, physically synthesize them, test their safety, and tweak their structures.
Because humans guide the process from start to finish and physically validate the results, the legal system considers them the true inventors. In this view, the AI acts as an incredibly advanced assistant or calculator, rather than an independent creator.
Marketing vs. Legal Reality
This creates a unique tension for leaders in the technology space. On one hand, companies want to highlight the power of their algorithms to investors, partners, and the public. Pointing out that an AI "discovered" a molecule demonstrates the power of modern software and attracts capital.
On the other hand, spending tens or hundreds of millions of dollars to take a drug through clinical trials is a huge financial risk. To justify that investment, companies need bulletproof legal protection. If a company fails to follow patent rules or lists an invention in a way that creates legal ambiguity, competitors could challenge the patent in court down the road. If successful, that multi-million dollar discovery could end up in the public domain for anyone to copy.
As a result, companies must walk a careful line: credit the technology in public relations and investor decks, but strictly attribute human ingenuity in official patent filings.
Looking Ahead
As generative AI models grow more capable, drawing the line between "human guidance" and "machine output" will become much trickier. What happens when an algorithm not only proposes a molecule, but also predicts its toxicity, designs the lab tests, and optimizes the clinical pathway with minimal human intervention?
For now, global intellectual property systems remain firmly anchored to human effort. Whether those rules will eventually evolve or whether we will need new frameworks to handle machine-assisted discovery remains an open question. What is clear today is that while AI can vastly speed up how we solve complex problems, the law still considers invention to be a uniquely human achievement.
Credits
Story reference and initial case details credited to reporting by MIT Technology Review.
Comments