When AI Designs a Drug, Who Gets the Credit?
Pharmaceutical drug discovery has historically been a human enterprise — decades of expertise, failed hypotheses, and iterative experimentation compressed into the rare compound that makes it to trial. AI systems are now performing substantive portions of that work: identifying targets, generating candidate molecules, predicting binding behavior, and filtering for toxicity at speeds no human team can match. The question of who receives legal and professional credit for these outputs is no longer hypothetical.
Patent law in most jurisdictions was written with a clear assumption: inventors are human. The United States Patent and Trademark Office and equivalent bodies in Europe and the UK have consistently held that AI systems cannot be named as inventors. But that position was formulated before AI moved from narrow pattern-matching into generative molecular design — before a system could propose a novel compound that no researcher had previously conceived.
The friction is already surfacing in active drug development pipelines. When an AI model generates a candidate molecule that becomes the basis of a patent application, the humans named as inventors are typically those who directed the system, curated the outputs, or refined the lead compound. Whether that attribution accurately reflects the epistemic contribution — and whether it holds up under legal scrutiny — is becoming a live dispute rather than an academic one.
The core legal problem is that inventorship carries weight beyond credit. It determines ownership stakes, licensing rights, and who can assign or commercialize a patent. If a pharmaceutical company's AI system produces the central inventive step in a drug candidate, and the named human inventors contributed primarily operational or supervisory work, the validity of that patent may be challengeable. Competing parties or generic manufacturers could argue that no qualifying human inventor existed — potentially rendering the patent unenforceable.
This creates a direct business risk. Companies building AI-first drug discovery pipelines are simultaneously building IP portfolios whose legal foundations may rest on an attribution framework that courts have not yet fully tested. The more capable and autonomous the AI system, the thinner the human inventive contribution becomes, and the more exposed that IP becomes to challenge.
There is also a subtler institutional problem. Scientific credit — authorship on papers, recognition in the research community, career advancement — runs through similar human-centric channels. When an AI system contributes meaningfully to a discovery, the humans who trained it, prompted it, or selected from its outputs receive the professional recognition. This is not inherently wrong, but it creates incentive distortions: teams may underreport AI contributions to preserve conventional authorship claims, which in turn obscures how much of the science was actually machine-generated.
Regulatory bodies have begun to notice. The FDA's evolving guidance on AI in drug development focuses on validation, transparency, and accountability — but does not yet address inventorship or authorship directly. The European Medicines Agency is in a similar position. The gap between what AI systems are doing in laboratories and what governance frameworks currently address is measurable and widening.
From an operational standpoint, pharmaceutical companies are navigating this without clear precedent. Some are adopting internal policies that document AI contributions in granular detail, partly to protect against future legal challenges and partly because regulatory submissions increasingly require disclosure of AI involvement. Others are treating the question as premature — a risk to manage later, when and if it becomes a formal legal issue.
That posture is unlikely to hold. As AI-designed compounds move through clinical trials and approach market approval, the inventorship question will be tested in adversarial contexts — licensing negotiations, patent litigation, generic drug challenges — where the answers carry direct financial consequences. The companies that have developed clear, defensible frameworks for AI contribution documentation will be better positioned than those that have not.
The deeper signal here is structural: AI is transitioning from a tool that assists human inventors to a system that originates inventive outputs. Legal and institutional infrastructure built for the former is not adequate for the latter, and the adjustment period carries real exposure.
Sources: — MIT Technology Review (https://www.technologyreview.com/2026/08/21/1142627/when-ai-designs-a-drug-who-gets-the-credit/)