Key takeaways
Licensing can be structured across different application areas, enabling multiple partnerships simultaneously.
Field-of-use licensing by geography or market segment can simultaneously serve public health goals in resource-constrained settings and commercial objectives in high-income markets.
Access to AI-enabled health innovations may be supported by IP strategies that take into account different models, for example, access models, including differential pricing commitments.
Dual strategies of licensing to others while developing internal products may provide more stability than single-approach models.
Cloud-based delivery approaches protect core technology while enabling broad deployment to health systems.
Emerging patterns in IP licensing strategies demonstrate how AI-enabled health innovators can support pathways that promote commercial viability and public health objectives.
2.1 Field-of-use licensing and parallel partnerships
A key lesson from AI-enabled health innovators is that licensing can be structured across different application areas or geographical territories, enabling multiple partnerships to operate simultaneously without conflict. This approach, known as field-of-use licensing, allows a patent holder to license the same underlying technology for different purposes, for example, licensing diagnostic AI algorithms separately for infectious disease detection, cancer screening and cardiovascular assessment.
By segmenting rights according to a therapeutic area, geography or market segment (such as high-income or low- and middle-income countries), innovators can partner with multiple organizations that are best positioned to serve different populations or address different health challenges. This maximizes both the reach of the technology and its impact, while generating diversified revenue streams that support continued R&D. From a global health perspective, the ability of innovators to structure licensing differently for diverse populations and markets is a significant tool. Field-of-use licensing enables tailored partnerships with local manufacturers or distributors in resource-constrained settings, facilitating technology transfer and capacity-building while preserving commercial opportunities in other markets.
2.2 Out-licensing and internal development
Strategies that combine out-licensing to external partners with internal product development may achieve broader health and commercial objectives than a single approach to licensing. Out-licensing may be necessary when the licensor lacks the resources to continue downstream development of IP, as this provides a pathway to valorizing it.
Companies that rely solely on internal commercialization may lack the resources or market access to achieve broad global impact. A flexible approach allowing both out-licensing and internal development allows innovators to maintain direct engagement with product development in selected markets, while leveraging partners' expertise, infrastructure and local knowledge in others. For technologies addressing global health challenges, dual strategies enable companies to pursue market commercialization that generates revenue to drive forward operations while simultaneously partnering with public health organizations or generic manufacturers for deployment in low-resource settings.
2.3 Cloud-based delivery as an access enabler
Innovative deployment models, particularly cloud-based or software-as-a-service architectures, demonstrate how technology companies can protect core IP while enabling broad deployment to health systems with varying technical capacities. By hosting AI algorithms, databases or analytical tools centrally and providing access through secure interfaces, innovators retain control over proprietary models and training data (often protected as trade secrets) while making functionality available to end users. This approach reduces barriers to adoption (as health facilities need not invest in specialized hardware or maintain complex software), while enabling continuous updates, quality monitoring and standardized performance across diverse settings.
From an access perspective, cloud-based delivery facilitates differential pricing commitments. Resource-constrained health systems pay reduced fees or access services through subsidized programs, while the same technology generates commercial revenue in higher-income markets. This model exemplifies how technical architecture choices intersect with IP strategy to advance both sustainability and access objectives. Yet cloud delivery also raises considerations around data sovereignty, cross border data flows, internet connectivity requirements and long-term service continuity as well as the environmental impacts of data servers. These issues could be managed and may require deliberate design choices, including offline-capable architectures, local data-hosting options and transparent service continuity commitments so that cloud-based models facilitate access.
2.4 Access models, including licensing
Access models may involve licensing agreements to support feasible real-world availability and benefits from innovations, in a manner that remains commercially efficient and sensitive to the conditions for downstream development and does not compromise safety, quality and efficacy. Access models include licensing on voluntary and mutually agreed terms, and allows adaptation to patient needs and the demands and unique attributes of a jurisdiction.
Access models may also include differential pricing commitments requiring lower prices in low- and middle-income countries, humanitarian provisions allowing royalty-free use in specified contexts, sublicensing rights that may support local production, technology transfer and capacity-building obligations, and transparency, all with the aim to improve access to medical countermeasures. Factors that may limit commitments to access models include limited facilities, size of market and innovator and expertise, which may impact downstream production and distribution.
For health technologies using AI, access models may mean structuring licenses to enable the adaptation of algorithms for local disease patterns, sharing training methodologies while protecting proprietary data sets or providing preferential licensing terms for public health applications. It may also encompass commitments to share model performance data across diverse population groups, ensure algorithmic non-discrimination, and support the local capacity to audit and validate AI tools. In addition to bilateral licensing, WHO and WIPO support IP licensing mechanisms, such as the Medicines Patent Pool (MPP), an independent public health organization founded by UNITAID, highlighting a balanced approach to the management of IP. The MPP enables patent holders to voluntarily license IP to enable the generic production of medical technologies, including in resource-limited settings.
These examples show how licensing approaches may embrace diversification and attention to social impact. For health technologies using AI, which often involve multiple layers of IP (such as algorithms, training data, software implementations and user interfaces), licensing structures can unbundle these elements, offering different terms for various components or use cases. By deploying IP strategically, in ways that advance both innovator and global health objectives, commercialization pathways can be closely intertwined with the goal of improved health outcomes for all.
Qure.ai, founded in 2016 in India, develops deep-learning algorithms for medical imaging and diagnostics. It has a particular focus on addressing health care gaps in low- and middle-income countries, where radiologist shortages and clinical backlogs limit access to timely diagnostic care. Its flagship product, qXR, detects abnormalities in chest X-rays, including tuberculosis (TB), lung nodules and pneumonia, while its expanding portfolio covers the AI-assisted interpretation of head CT scans for stroke triage, guidance for placing breathing tubes and other emergency care applications. The company's tools have been deployed in over 90 countries across Africa, the Americas, Asia and Europe, serving approximately 15 million patients annually. TB screening alone analyses up to 10 million chest X-rays per year.
Qure.ai's competitive position rests primarily on regulatory clearances, proprietary training data sets and algorithmic implementations developed specifically for resource-constrained settings, rather than on a prominent patent portfolio. Its most strategically valuable confidential assets are likely its curated imaging data sets, drawn from diverse patient populations underrepresented in Western training data; technical implementations addressing challenges common in low- and middle-income countries, such as low-quality equipment and high TB prevalence; and operational expertise in integrating AI tools into resource-limited clinical workflows and mobile screening programs.
In low- and middle-income countries, regulatory clearances from recognized authorities provide the independent clinical validation that health ministries and international procurement bodies require to justify deployment at a national scale. For its chest X-ray TB solution for use without a human reader, Qure.ai had secured 18 FDA 510(k) clearances as of late 2024, Class IIb CE Mark certification under the EU Medical Device Regulation and WHO policy recognition of its chest X-ray TB solution following independent evaluation demonstrating that it met WHO performance standards for tuberculosis screening. The company has raised approximately USD 125 million from investors, including Peak XV Partners, Novo Holdings and Lightspeed India Partners, as well as a USD 8 million grant from the Gates Foundation, supporting continued expansion across high-, middle- and low-income country markets.
Key lessons for innovators
Leveraging local health care needs and vast imaging data sets can provide an advantage. Health-tech innovators in low- and middle-income countries can deliver globally relevant diagnostic AI solutions, achieve iterative regulatory clearances across multiple jurisdictions and secure funding from strategic global investors.
Regulatory validation (FDA, CE mark) combined with WHO normative guidance provides market credibility.
Public health partnerships enable large-scale validation and deployment that would be costly for commercial-only strategies.
The role of mixed funding points to sources outside private venture capital. These include public health program partnerships, national government procurement and multilateral and bilateral health funding.
BenevolentAI is a UK-based AI company that applies machine learning and large-scale biomedical knowledge graphs to drug discovery and target identification. It integrates structured biological data, scientific literature and experimental evidence to uncover previously unrecognized disease mechanisms and therapeutic targets. Since 2019, the company has collaborated with AstraZeneca across multiple disease areas, including chronic kidney disease, idiopathic pulmonary fibrosis, systemic lupus erythematosus and heart failure. It has identified novel targets that have progressed into AstraZeneca's research portfolio and demonstrated the commercial viability of AI-enabled target discovery when incorporated into established pharmaceutical pipelines.
BenevolentAI's IP strategy combines patents protecting core platform methods with trade secret protection for its most competitively sensitive assets. Representative patent filings, including WO 2024236317A1 and US 20250022615A1, describe systems for identifying and ranking therapeutic targets by integrating machine learning inference with the structured querying of biomedical knowledge graphs. Claims are framed around applied, system-level implementation, rather than abstract algorithms. This satisfies the technical requirements of the EPO and the inventive concept requirements of the Alice/Mayo framework in the United States. It also provides protection against design-arounds by competitors using different machine learning models within similar knowledge graph architectures.
To complement these patents, the company maintains its most valuable assets as trade secrets, including a proprietary biomedical knowledge graph drawing on over 40 million scientific papers, curated training data sets recording drug target trial outcomes, trained model weights and target prioritization scoring functions. This patents-for-methods and trade-secrets-for-data approach reflects a deliberate recognition that methods are at risk of reverse engineering through publications. Data curation and model training are substantially harder to replicate.
BenevolentAI was listed on Euronext Amsterdam in April 2022 through a special purpose acquisition corporation merger valued at approximately USD 2 billion. Its collaboration agreement with AstraZeneca is structured to allow each party to retain background IP while jointly owning novel targets identified through the collaboration. AstraZeneca holds exclusive licensing rights in agreed therapeutic areas; BenevolentAI retains platform rights for use in other partnerships, generating revenue through a combination of research funding, milestone payments and downstream royalties.
Key lessons for innovators
A strategic IP allocation can maximize protection without disclosing core competitive advantages. BenevolentAI illustrates the value of deliberately deciding which innovations to patent and which to keep confidential. Patents cover novel computational methods, system integration and algorithmic innovations that competitors could replicate. Trade secrets protect training data, knowledge graph content and model weights, and target prioritization heuristics that cannot be reverse engineered.
Partnership IP structuring can support field-of-use exclusivity. In its deal with AstraZeneca, BenevolentAI granted exclusive rights in specific therapeutic areas (chronic kidney disease, idiopathic pulmonary fibrosis, systemic lupus erythematosus and heart failure) while retaining platform rights for other indications. This enables multiple disease-specific partnerships using the same technology, increasing platform value. For AI–pharma collaborations, therapeutic area exclusivity is preferable to platform-wide exclusivity, as it supports diversified revenue streams.
Drafting claims that emphasize architectures, data structure integration and technical improvements strengthens long-term enforceability. BenevolentAI's patents focus on integrated system architectures combining knowledge graphs and machine learning. This strategy has been effective in systems in the United Kingdom/EPO and United States that allow functional claims with structural support. Such claims are broader, more defensible against design-arounds and more likely to survive eligibility challenges than narrow algorithmic claims.
A hybrid business model combining a platform and pipeline diversifies revenue and reduces reliance on any single source. The company mitigates risk by pairing platform licensing, exemplified by AstraZeneca's access to its AI system, with internal drug development programs, such as BEN-2293 and BEN-8744.
Regulatory exclusivities complement IP. BenevolentAI's clinical candidates will gain regulatory data exclusivity in addition to patent protection. Combining composition-of-matter, method-of-use and formulation patents with regulatory exclusivities can extend effective market protection to 20 to 25 years, from initial discovery through generic entry.
CytoReason, founded in 2016 and headquartered in Tel Aviv, applies machine learning to large-scale, harmonized, multiomics data sets to reconstruct cell-level representations of immune-mediated disease. Its computational disease models, sometimes described as digital twins, generate interpretable immune cell interaction maps and disease state models. Pharmaceutical partners use these to support target discovery, indication selection, patient stratification and the prediction of clinical responses across immunology and inflammation-driven conditions.
The company's IP strategy combines patent assets with extensive trade secret protection. Its long-term competitive advantage derives primarily from the scale, quality and continuous refinement of its proprietary biological data sets and disease models rather than its patent portfolio alone. CytoReason operates a cloud-based business-to-business platform through which pharma partners submit analytical queries via API and receive outputs such as target recommendations and biomarker candidates. Underlying data sets and model architectures remain confidential and protected from reverse engineering. This architecture is reinforced by contractual provisions granting partners exclusive rights to specific targets or biomarkers within defined therapeutic areas. CytoReason retains ownership of the core platform and the freedom to collaborate with other partners across different disease domains, enabling multiple simultaneous partnerships without conflict.
CytoReason has established a strong portfolio of pharmaceutical partnerships, reflecting the commercial traction of this model. Its collaboration with Pfizer, expanded in 2022, has a total potential value exceeding USD 110 million and includes an equity investment. The company has also announced collaborations with Sanofi, Ferring Pharmaceuticals, Merck KGaA and other biopharmaceutical organizations, applying its disease models across immunology, inflammation, oncology and reproductive health programs.
Key lessons for innovators
A business-to-business pharma platform model offers an alternative to clinical deployment. Health companies using AI can build sustainable businesses by licensing to pharma R&D teams rather than deploying clinical decision support tools.
Trade secrets provide protection that is superior to patents for data-driven platform where a competitive advantage derives from accumulated data assets rather than novel algorithms.
Field-of-use exclusivity enables multi-partner models. CytoReason partners with multiple competing pharma companies simultaneously by granting each partner exclusive rights in specific therapeutic areas while retaining platform ownership.
A cloud platform architecture protects trade secrets. CytoReason's API-based platform delivers analytical outputs without exposing underlying data or models, enabling monetization while maintaining trade secret protection. This contrasts with on-premise deployments where customers could reverse engineer proprietary methods.