Insilico's AI found a drug, but the patent names only humans
When the biotech company Insilico Medicine used its computer models to propose a promising drug candidate for pulmonary fibrosis, it said in a press release that its generative AI platform had discovered the molecule. But when Insilico filed for the patent that protects the compound's chemical structure, the filing made no mention of AI: it names five humans, including CEO Alex Zhavoronkov, as the invention's legal inventors.
The gap traces back to a 2022 ruling. Patent attorney Ryan Abbott, a partner at the Los Angeles firm Brown, Neri, Smith & Khan, brought a pro bono test case arguing that an AI system called DABUS should be named the inventor of a food container whose faceted surface improves heat transfer and stacking, since no human had contributed to that design. An appeals court in Washington, DC, disagreed. It held that US patent statutes define an inventor as an "individual," and the plain meaning of that word is a human being; the court called questions about whether an AI deserves legal rights "metaphysical matters" beside the point.
Sarah Korman, a patent attorney who is now chief business officer and legal officer at Isomorphic Labs, Alphabet's AI drug discovery spinout, says the rule is settled: "There needs to be a human inventor or there's no invention and no patent." She adds that the law will still need to evolve to keep up with AI. The US Patent and Trademark Office has itself acknowledged that an AI system, like other tools, may perform acts that would count as inventorship if a human did them, but under the current approach it treats AI as a mere tool, like a calculator, that need not even be mentioned in a filing. That reverses Biden-era USPTO guidance that had helped applicants work out when a human should count as a co-inventor of an AI-assisted discovery; the agency changed course after the Trump administration arrived.
Abbott warns the ambiguity cuts both ways. Because one way to invalidate a patent is to show it lists the wrong inventors, future litigants could challenge AI-generated drug patents on exactly that basis. He also worries that if US policy ends up excluding AI-generated outputs from protection altogether, it could dampen drug development, since intellectual-property rights exist, per Article 1 of the US Constitution, to "promote the Progress of Science and the Useful Arts." A parallel fight is already under way in copyright: the US Copyright Office is refusing to grant copyright to AI-generated images and text, a stance that has drawn objections from groups including the Motion Picture Association of America.
For now, Zhavoronkov says Insilico keeps humans in the loop and documents their contributions carefully: chemists still have to synthesize candidate drugs, create variants and test them on animals, and "that's the person who is going to be named on the patent." He adds that even a fully roboticized pipeline would still need someone to push the button and fund the budget. Whether pushing that button is enough to count as inventing remains untested. Abbott poses the sharper version of the question: "What if I asked Claude to cure cancer, and it did? I think it would be inappropriate to claim that I invented that."
Key facts
- Insilico Medicine's press release credited its generative AI platform with discovering a pulmonary-fibrosis drug candidate, but the patent for the molecule names five human inventors, including CEO Alex Zhavoronkov, not the AI.
- A 2022 Washington, DC appeals court ruling, from a pro bono test case Ryan Abbott brought naming the AI system DABUS as inventor of a food container, held that US patent law's 'individual' inventor requirement means a human being.
- Sarah Korman, chief business officer and legal officer of Isomorphic Labs, says a human inventor is required or there is no patent, though she expects the law to evolve to keep pace with AI.
- The USPTO now treats AI as a tool like a calculator that need not be mentioned in filings, reversing Biden-era guidance on when AI use makes a person a co-inventor, after the Trump administration arrived.
- Abbott warns AI-generated drug patents could face legal challenges over improperly listed inventors, and that excluding AI outputs from IP protection could dampen drug development.
Why it matters
AI models can now generate atomic-level drug designs almost as easily as a chatbot writes a thank-you note, but patent law has not moved to match. The Insilico case shows the resulting gap in plain view: a company happy to credit its AI publicly still has to reconstruct a human inventor story for the patent office, because US courts have settled that only a human, not a machine, can hold inventor status. That mismatch shapes how every AI-driven drug discovery company now has to operate, document, and litigate.
Who it affects
AI drug discovery companies such as Insilico Medicine and Isomorphic Labs, whose patents are the commercial payoff of their models; patent attorneys and litigants who now have a clear legal test to apply or challenge; and the US Patent and Trademark Office, whose guidance on AI and inventorship has already reversed once with a change of administration. The same question extends beyond patents: the US Copyright Office's refusal to grant copyright to AI-generated images and text raises the identical issue for another branch of IP law, drawing objections from groups like the Motion Picture Association of America.
How to use it
Companies building patentable work with AI assistance need a documented human contribution to list as inventorship, not just a model output. Insilico's approach, per Zhavoronkov, is to keep chemists synthesizing candidate compounds, creating variants and running animal tests, so there is always a person whose work supports being named on the patent, even as the ideation itself comes from the AI platform.
How solid is it
The article rests on a named, on-record 2022 appellate ruling in the DABUS case, direct quotes from the attorney who brought that case (Ryan Abbott), a named Insilico executive (Alex Zhavoronkov), and a named former patent attorney now at Isomorphic Labs (Sarah Korman), plus the USPTO's own public statements about AI and inventorship. It does not give exact dates for when Insilico filed the pulmonary-fibrosis patent or when USPTO guidance was published and reversed, and it does not name the other four human inventors on Insilico's patent.
Risks and caveats
The legal line is not fully settled: it is untested how much human contribution is enough to qualify as inventorship, and Abbott expects future litigation to test exactly that question against AI-generated drug patents. USPTO policy has already reversed once with a change of administration, so the current don't-ask-don't-tell treatment of AI as 'just a tool' could shift again. The parallel refusal of AI-generated copyright suggests the same uncertainty could spread to other categories of AI output before the law catches up.
“What if I asked Claude to cure cancer, and it did? I think it would be inappropriate to claim that I invented that.”
— Ryan Abbott, patent attorney