AI-Generated Active Ingredients: Who Owns the Invention?

The Technology Is Here. The Law Faces New Challenges.

Insilico Medicine brought a drug candidate for idiopathic pulmonary fibrosis from conception to Phase I trial in under 30 months – a process that traditionally takes four to six years. The company is now in Phase II. Recursion Pharmaceuticals, merged with Exscientia under one roof since late 2024, is advancing multiple AI-powered candidates in oncology and rare diseases.

These successes mark a turning point. AI in drug discovery is no longer a pilot project. And this raises a question that long sounded academic but now has immediate practical significance: Who holds the rights to AI-assisted inventions?

What BGH, EPO, and USPTO Have Decided – and What That Means

The current legal position is clear, even if inconvenient: AI cannot be an inventor.

The European Patent Office (EPO) established this in 2020 in the well-known DABUS case and has consistently maintained that position since. The German Federal Court of Justice (“BGH”) confirmed in its decision of June 11, 2024 (Case No. X ZB 5/22) that only a natural person can be an inventor – AI can at most be mentioned as a tool. The United States Patent and Trademark Office (“USPTO”) also followed suit in 2024 with updated guidelines.

What sounds like a purely formal hurdle at first glance has significant practical consequences:

• The human contribution must be documentable: Companies using AI tools in drug discovery must be able to demonstrate that a human made a substantial inventive contribution – in target selection, interpretation of results, or further development of the candidate. Mere “prompting” of an AI is not sufficient.
• Gray areas remain: How substantial the human contribution must be in individual cases has not yet been definitively clarified. This creates points of attack – for competitors seeking to challenge a patent, or for licensing partners questioning its validity.
• Internal documentation becomes mandatory: Development processes must be documented in such a way that the human contribution to an invention can be traced at any time in case of dispute. This is not an academic exercise, but a prerequisite for patentability.

License Agreements and AI-Generated IP: Where Classic Structures No Longer Fit

Even when an AI-assisted active ingredient is successfully patented, new legal questions arise at the next step. Classic license and partnering agreements were not designed for a world in which a machine has a substantial share in the invention.

Three areas deserve particular attention:

• Definition of licensed IP: Classic license agreements define licensed IP through patent lists or know-how descriptions. AI-generated IP often also includes the underlying training data, models, and algorithms. Those who do not explicitly include or clearly exclude these open up room for interpretation – and thus potential for conflict.
• Ownership in co-development: When two companies jointly conduct AI-assisted research and a candidate emerges, the question arises: Who owns the result? Classic joint IP clauses often do not fit constellations in which one party provides the AI platform and the other contributes biological or medical expertise.
• Contracts with AI providers: This is the biggest blind spot. Many pharmaceutical and biotech companies use AI platforms from third-party providers – whether as a SaaS solution or within a research collaboration. What is stated in the terms of use or in the research agreement has a decisive influence on the economic allocation of results as well as on existing usage and exploitation rights. This contract is often the most IP-critical in the entire process – and is the least carefully negotiated.

What Companies Should Do Today

The legal landscape surrounding AI-generated IP will evolve dynamically over the coming years. Companies that act now have a clear advantage over those that wait.

Three actions can be implemented immediately:

• Document development processes: Internal workflows should be designed so that the human contribution to the invention is recorded in a traceable manner at every AI-assisted step – as a basis for subsequent patent applications and to defend intellectual property rights in case of dispute.
• Review contracts with AI providers: Anyone using an external AI platform should clarify before signing the contract who owns the generated results, models, and data – and have this explicitly regulated in the contract.
• Review existing license agreements for currency: Companies that have introduced AI-assisted research after contract signing should examine whether existing IP definitions and rights grants still adequately reflect the new technical reality – particularly in co-development constellations.

Conclusion

AI is changing how active ingredients are discovered. This is no longer a prediction – it is reality. The legal and contractual consequences affect companies using AI tools today, right now. Those who do not adapt IP strategy and contract structures risk undermining the value of their own research.

Rödl supports pharmaceutical and biotech companies in the legal and tax structuring of AI-assisted development projects – from IP strategy through contract design with AI providers to safeguarding in license and partnering deals.