The AI Patent Paradox: Who Owns an Algorithmically Designed Drug?

As generative AI evolves from writing essays to designing complex molecular structures, a massive legal gray area is emerging in the biotech sector. While companies like Insilico Medicine claim their AI platforms "discover" new drugs, the legal framework for intellectual property remains stubbornly human-centric.

The Discrepancy Between Discovery and Inventorship

In a recent breakthrough for pulmonary fibrosis, the biotech firm Insilico Medicine utilized its generative AI platform to propose a promising new molecule. While the company's marketing materials celebrated the "AI-discovered" nature of the molecule, the subsequent patent filing told a different story. The patent named five humans, including CEO Alex Zhavoronkov, as the inventors, omitting any mention of the AI.

This discrepancy is not a clerical error but a strategic necessity dictated by current law. As AI models become capable of generating atomic designs as easily as ChatGPT generates text, the industry faces a critical question: if an AI does the heavy lifting, does the discovery even qualify for patent protection?

The current legal impasse stems from landmark court rulings, specifically the case involving DABUS, an AI system used to design an optimized food container. Ryan Abbott, a partner at Brown, Neri, Smith & Khan, argued that because the AI performed the inventive steps without human intervention, it should be named the inventor.

However, in 2022, a Washington, D.C., appeals court dismissed the argument. The court ruled that US statutes define an inventor as an "individual," a term the court interpreted to mean a human being. This "metaphysical" distinction creates a massive hurdle for AI-driven innovation: without a human inventor, there is no invention, and without an invention, there is no patent.

The Risk to Global Innovation

This legal rigidity poses a significant threat to the pace of scientific advancement. Industry experts like Sarah Korman of Isomorphic Labs (an Alphabet spinout) argue that while AI is undeniably inventing things, the law must evolve to keep pace.

The stakes are particularly high in the pharmaceutical industry. If AI-generated outputs are excluded from patent protection, companies may lose the incentive to invest the billions required for clinical trials. Furthermore, legal experts warn of a "wrongful inventorship" loophole; if a patent is filed claiming human inventorship when an AI actually performed the core creative task, the patent could be challenged and invalidated.

The "Don't Ask, Don't Tell" Era of AI Patents

To navigate this, the US Patent and Trademark Office (USPTO) has shifted toward a pragmatic, albeit controversial, approach. While previous guidance sought to help applicants define human contribution, the current stance treats AI merely as a tool—no different from a calculator.

For now, companies like Insilico are maintaining "humans in the loop" to secure their IP. By ensuring human chemists are involved in synthesizing variants and conducting animal testing, they provide the legal "hook" necessary to name a human inventor. As the technology approaches full autonomy, the industry awaits a new legal definition of what it means to "invent."

Key Takeaways

  • Legal Definition of Inventorship: Current US law requires an "individual" to be a human, meaning AI-only discoveries currently cannot be patented.
  • Strategic Compliance: Biotech firms are intentionally involving humans in the synthesis and testing phases to ensure they meet the legal criteria for human inventorship.
  • Innovation Risk: If patent laws do not evolve to recognize AI-assisted breakthroughs, the massive capital investments required for drug development may be at risk.