The Patent Paradox Threatening Billions in AI-Discovered Drugs

The race to harness artificial intelligence for drug discovery has produced a legal collision course. Biotech companies are publicly celebrating AI platforms as the inventors of promising new drug candidates, yet when filing patent applications, they must name human employees as the creators—a requirement that raises fundamental questions about whether these patents will withstand future legal challenges.

The tension came into sharp focus when Insilico Medicine developed a pulmonary fibrosis drug candidate using generative AI. Despite crediting their AI platform with the discovery, the company's patent application names five humans, including CEO Alex Zhavoronkov, as inventors. This pattern is becoming standard across the AI drug discovery sector, where companies navigate an increasingly precarious gap between technological reality and legal requirements.

The Human-Only Rule and Its Discontents

U.S. patent law unambiguously requires human inventors. A 2022 ruling by a federal appeals court in Washington, DC, established that "individual" in patent statutes means human being, decisively rejecting a test case filed by attorney Ryan Abbott that attempted to name an AI system called DABUS as inventor of a food container design.

The U.S. Patent and Trademark Office has oscillated on how to handle AI's role in invention. Under the Biden administration, the office published guidance on AI co-inventorship, but the Trump administration reversed course, now treating AI as a tool comparable to a calculator that requires no special disclosure on patent applications. This "don't ask, don't tell" approach may provide short-term convenience but creates long-term vulnerability.

The stakes extend beyond individual patents. If the wrong inventors are listed on patent applications—a risk when AI generates novel compounds with minimal human contribution—those patents could be invalidated. For pharmaceutical companies, such invalidation could erase billions in exclusivity rights and undermine the business model funding AI drug development.

When the Law Lags Behind Innovation

Sarah Korman, Chief Business Officer and Legal Officer at Isomorphic Labs, an Alphabet spinout focused on AI drug discovery, acknowledged that laws will need to evolve to keep pace with AI's expanding role. The current framework, rooted in Article 1 of the U.S. Constitution's mandate to grant exclusive rights to inventors "to promote the Progress of Science and the Useful Arts," was designed for an era when invention was inherently human.

The U.S. Copyright Office has taken an even harder line, refusing copyrights to AI-generated images and text. This broader pattern suggests American intellectual property law is struggling to accommodate machine creativity, even as that creativity accelerates in pharmaceutical applications where speed could mean earlier access to life-saving treatments.

For AI drug discovery companies, the human-inventor requirement creates operational challenges. Firms must maintain detailed documentation of human involvement throughout the discovery process to secure patent protection, potentially slowing automation and increasing costs. Companies face a strategic tension between marketing their AI capabilities to investors and ensuring sufficient human contribution for legally defensible patents.

The Competitive Question

As AI capabilities advance and the gap between machine contribution and human involvement widens, the current legal framework may create competitive disadvantages for U.S. companies versus jurisdictions that adapt their intellectual property laws. The question facing policymakers is whether the human-inventor requirement will accelerate or hinder the AI drug discovery market's ability to deliver faster, cheaper therapeutics—and whether the constitutional goal of promoting scientific progress is better served by maintaining traditional inventorship concepts or evolving them for the age of artificial intelligence.

Sources

1. When AI designs a drug, who gets the credit? — MIT Technology Review