AI use by life sciences companies – strategic introduction

Technology background and building blocks

Contrary to what popular culture and mass media in recent years might lead us to conclude, Artificial Intelligence (AI) has been a focus of academic interest and research for decades, reaching back to at least 1950 when Alan Turing more formally asked “Can machines think?”.[1] In 1956, John McCarthy is said to have coined the term “artificial intelligence” at a conference or workshop.[2] Over the years that followed, computing power and storage have advanced, becoming more widely available and cost-effective. As a result, commercial and even consumer AI applications are now more common.

At a high level, it can help to distinguish generative from non-generative AI. Generative AI typically produces new content in response to a user’s written query, e.g., outputting an image, music or text resembling a creative work. In contrast, non-generative AI typically produces an answer or insight based on analysis of a query data submission, e.g., whether or not the cells seen in a radiology image are cancerous.

Drilling down into the technology involved in a non-generative AI platform, primary elements often include (i) training data used to create an AI model, (ii) software used to both analyze training data and create the model and later use such model to analyze query data to provide an output, and (iii) the AI model created using such training data and software. Further iterations and additional models are also possible, and data annotation and fine-tuning of software is often required. As a technical matter, each of the three elements above might be internally developed or sourced, obtained from a third party, or a hybrid of the two (e.g., a combination or customization of the foregoing). The following are a few elements to bear in mind in considering such alternatives as part of an overall IP strategy.

Platform or output use and materiality

In scenarios where the company intends to distribute or otherwise commercialize or allow third parties to use the AI platform it is implementing or developing, or parts thereof, there are several considerations that may be key while being less central in other scenarios. For example, it is often important in this case to analyze the expected scope of company’s ownership, exclusive rights and other proprietary rights relating to the AI software, training data and models. Copyrights in software and potentially in data, database rights, patents and trade secrets may all be implicated. Applicable laws may depend on the locations of the various parties or sources of technology. Of note for companies that are experienced with patents but less so with copyrights, the required contribution to rise to the level of authorship or joint authorship under applicable copyright law can vary and depend on the facts, and is different from that required for inventorship or joint inventorship under patent law. As a result, having appropriate written agreements in place with parties that participate in or otherwise contribute to the company’s AI development, implementation or use are often important sources of some certainty as to the company’s rights. In addition, the terms and conditions of any licenses, terms of use or other agreements that apply to any third party components of the AI platform can also be material factors in the company’s rights.  

In other situations, including where the company will use the AI platform internally and leverage the output as part of its overall R&D activities, the identity and materiality of the specific output in relation to the company’s own IP portfolio and proprietary asset development should be considered. As further context, it is worth noting that the United States Patent and Trademark Office issued guidance earlier this year on patentability and inventorship for AI-assisted inventions and the U.S. Copyright Office is expected to issue a report in the coming months analyzing the impact of AI on copyright and making recommendations about legislative or regulatory action in light of its prior position that U.S. copyright protection should not extend to AI-generated works. As a result, the scope of intellectual property protection for AI-related developments is in flux to some extent in the U.S., as may also be the case in other jurisdictions, reinforcing the importance of working with counsel that is tracking and analyzing changes or other updates relating to applicable laws as they occur.

Next steps

For additional information and resources regarding AI, see Hogan Lovells’ AI Hub. Please contact the author or the Hogan Lovells attorneys with whom you regularly work to discuss your specific needs with respect to developing strategies around AI development, implementation or use.

This is an article in our “Life Sciences Transactional Insights” series, which aims to provide key practical takeaways for our transactional colleagues by anticipating the needs of their regulatory, intellectual property, and business stakeholders. Our dedicated team of life sciences and health care licensing and commercial transactions lawyers understand the challenges and opportunities that strategic alliances and other partnering relationships present. We draw on the depth of our life sciences practice and work seamlessly with our regulatory experts to provide unparalleled transactional support. Ensure you are subscribed to Hogan Lovells Engage to receive our insights.