Looking forward to the 2025 prediction season, a survey of the emerging AI-powered drug discovery landscape has been making the rounds—and with good reason. The survey, summarized by Stanford’s Chris Bradbury, is likely the most comprehensive public review of advancements in this rapidly evolving field to date.
AI-powered drug discovery is categorized into four distinct waves of development, providing a clear framework for understanding how this technology has progressed. The most recent wave, in particular, emphasizes two key trends: (1) leveraging more data and extracting more value from diverse data sources through multimodal, multiscale, synthetic, and self-supervised approaches; and (2) focusing on more specialized pipelines, where feedback from the development process may be used more effectively.
The report also frames current and potential business models and value propositions that could shape the future of AI drug discovery, synthesizing the following trends for the most recent wave of companies: they are more specialized, unlocking therapeutically relevant capabilities to focus on specific diseases or therapeutic areas; intend to reach clinical stage faster (reflecting the goal of a more focused pipeline); and emphasize higher value indications and the economic potential of real-world clinical assets.
Key questions raised by these insights include whether greater speed to clinical stage could facilitate faster data collection, leading to a feedback loop that accelerates innovation. Another question is whether there is inherent tension between end-to-end approaches and federated models, with players in model-as-a-service, data supply, and lab automation potentially emerging as dominant forces.
Given this highly dynamic landscape, the AI drug discovery sector will require commensurate innovation in legal/risk management frameworks to ensure that (1) the value of the AI technologies are being protected and realized, (2) proper governance is in place around critical data and technology, and (3) these goals are advanced in step with the most fundamental goal of the field: bringing more effective drugs to more patients at lower cost.
Key pillars of such a framework include:
Intellectual Property
Our earlier review of AI drug discovery patenting found that the landscape is wide open. An AI drug discovery company that is pursuing a sophisticated IP strategy should consider what parts of the tech stack they derive value from, and take advantage of the open landscape to build their patent portfolio around those key parts. For example, model developers should consider whether there is a critical aspect to their architecture that competitors would seek to replicate, including aspects that allow for platforms to be flexibly deployed for various types of targets and therapeutic areas. (Wet) lab automation companies should continue to pursue medical device-type strategies around the structure and functions of their devices, but also ensure they are protecting any computer vision or other sensor data processing as well as how final outputs from their proprietary technologies are processed for delivery to customers.
For all companies, this will allow for more effective negotiations in licensing and partnership deals, or in demonstrating defensibility to investors.
More broadly, a robust IP strategy should balance resources allocated to patents and trade secrets. Similarly, executing an IP strategy can require spending on the most critical aspects of the technology and business model, while deferring costs for other aspects. As such, companies whose value is primarily in proprietary technologies might devote more resources to a dense patent portfolio backstopped by trade secret protection. Companies whose value is around collaboration, data sharing, and/or strategic partnerships may consider focused patent filings sufficient to protect foundational technologies to support licensing and other collaborative efforts, while relying on copyright protection and confidentiality provisions to bolster protection.
Privacy, Confidentiality, and Cybersecurity
These data protection considerations have heightened importance in the drug discovery context, due to the criticality of the data being managed as well as the potential overlap with patient/medical data.
As a reference point, most LLM platforms and developers building on top of LLMs have matured in how they manage risks in these key areas on behalf of their customers. For example, most companies that receive data from customers/users are touting their security certifications (e.g., SOC2, etc.), responsibility and transparency around data usage and model training (i.e., not training models with customer data under commercial licenses; taking no ownership in model outputs), and enforcement of confidentiality provisions both internally and with subprocessors.
Firms in the AI drug discovery space on the model developer side or customer/user side, respectively, should provide or seek (e.g., during diligence or licensing negotiations), similarly robust standards. In addition, they should take extra caution to ensure appropriate consents for current and future uses of any data subject to privacy obligations. This will become critical as more data from clinical trials becomes integrated into the AI drug discovery pipeline. The newest wave of AI companies focused on highly relevant data generation, for example, may seek more stringent contractual and operational protections for this data.
FDA and Other Regulatory Considerations
The FDA is expected to release draft guidance (originally estimated by end of 2024) relating to the use of AI in drug development. An earlier publication emphasized evaluating risks of AI in drug development based on the specific context of use. Key factors highlighted as part of such evaluation include trustworthy and ethical AI, managing bias, quality of data, and model development, performance, monitoring, and validation.
In view of these factors, firms seeking to be proactive about potential regulations around AI in drug discovery should define their risk profile, and based on the risk profile, identify processes and controls to address identified risks. Firms may need to place outsized emphasis on validation, for example, where aspects of the drug evaluation process are at least partially substituted with AI models.
Summary
Taken together, these major legal considerations for AI drug discovery firms point to a legal strategy built around:
- Defining (a) what their value proposition is; (b) what data, technology, and talent support the value proposition; and (c) what use case-specific risks may be in play.
- Strategically allocating resources across various IP assets.
- Creating an AI/data governance team and framework that aligns policy with specific controls for the identified risks.
- Executing on these foundations throughout model development, testing, and deployment, management of sensitive data, licensing and partnership negotiations, and diligence for investment and/or M&A.
By establishing and executing on this framework, the next generation of AI drug discovery firms can continue to focus on their differentiation in data, technology, and targets, while being well positioned for licensing and investment.
Foley will continue to monitor these developments. Foley is here to help you address the short and long-term impacts in the wake of regulatory changes. We have the resources to help you navigate these and other important legal considerations related to business operations and industry-specific issues. Please reach out to the authors, your Foley relationship partner, or to our Artificial Intelligence Are of Focus and Health Care & Life Sciences Sector with any questions.
