The emergence of AI-generated “digital twin” technology presents interesting new applications for life sciences businesses, including the simulation of patient responses, improvements in clinical trial efficiencies and new approaches for evidence generation. As adoption of this developing technology grows, so too does the need to understand the intellectual property considerations surrounding it.
Use of digital twins in a clinical context
In the clinical setting, a digital twin is a dynamic, virtual representation of a patient or patient cohort, connected to its real-world counterpart by a two-way flow of real-time data. Built using AI models trained on extensive biological and clinical datasets, digital twins can replicate disease progression, predict treatment responses, or act as virtual comparators within clinical trials. This in turn allows organisations to improve efficiency and accelerate innovation with data-driven decisions.
Pharmaceutical companies are currently exploring the use of digital twins in a variety of applications, including to run in silico simulations before initiating human studies, reduce the size of control groups, and optimise dosing and treatment pathways in precision medicine. Although not yet widely used in formal submissions, organisations also are increasingly monitoring how modelling and simulation‑based evidence may be incorporated into future regulatory pathways.
Who owns the IP rights in a digital twin?
Digital twins sit at the intersection of algorithm development, data science and biomedical analysis. As a result, IP ownership is rarely straightforward, and what appears to be a single output is often the product of multiple overlapping contributions from different parties. The fact that digital twins evolve over time, refining themselves through new data inputs, adds an additional layer of complexity to determining ownership and control.
Key questions in this area include:
- Who owns the trained AI model – the technology provider, the trial sponsor, or both?
- What rights apply to the underlying datasets, particularly when patient‑derived information is involved?
- Who controls the improvements generated as the model is updated, retrained or adapted for new therapeutic contexts?
To address these issues, any collaboration and licensing agreements in this area must contain robust, well-drafted provisions governing ownership of the base model and any derivatives, rights in training data and outputs, and improvements to the system generated over time.
In practice, many of these questions will turn on the specific technical architecture of the model and the commercial arrangement between the parties, so early specialist input is advised. As always, careful contract drafting at the outset of any venture or partnership is essential to minimise the risk of disputes and to ensure each party’s freedom to operate as expected.
Data rights and confidentiality
Because digital twins depend on diverse datasets, organisations must consider data rights and confidentiality from the outset of any such project. Unclear data provenance or insufficient usage rights can expose companies to significant contractual, regulatory and IP risks, and the use of sensitive health data carries additional safeguarding requirements. Similarly, sharing models or outputs with partners must be handled carefully to avoid inadvertently disclosing proprietary algorithms or know‑how.
Patentability
Many patent systems (including the UK and Europe) are cautious about granting patents for inventions that are seen as purely abstract, such as where claims are directed solely to algorithms or mathematical models. As a result, an application presenting a digital twin purely as a predictive model or algorithmic output may struggle to meet patent eligibility requirements.
In practice however, digital twin technologies often do more than just process data in the abstract, and many involve concrete technical steps such as the physiological modelling of disease progression or the simulation of treatment responses. Framing these elements effectively at the application stage can therefore support their patentability.
For innovators, successful strategies often involve:
- Emphasising the technical effect or biomedical application of the technology – in other words, showing that it solves a technical problem in a technical way, rather than merely generating data.
- Claiming the specific architecture or training method, rather than broad conceptual claims.
- Linking the digital twin to a real-world medical device or diagnostic workflow.
A targeted patent strategy can help innovators secure protection for their inventions while avoiding excluded subject matter, but the margin for error is small. Minor changes in claim wording can determine whether a digital twin invention is seen as a patentable technical solution or an unpatentable abstract concept.
In such a complex and highly regulated area, early input from specialist patent counsel is always advisable, particularly to ensure that any filings support an organisation’s broader development and commercialisation plans.
Conclusion
Digital twins present a valuable opportunity for innovation in clinical development, but they also raise nuanced IP considerations. Businesses that proactively address ownership, patentability, and data rights will be well positioned to take advantage of this promising technology as it continues to mature.
