Regulation of Artificial Intelligence in the Pharmaceutical Sector

The implementation of AI as a universal shift toward data-centric drug discovery started in earnest as far back as 2012, with 2019 widely seen as a breakthrough year with Deep Genomic’s AI driven platform for predictive oligonucleotide targeting, as well as the launch of Google DeepMind’s ground-breaking protein structure modelling engine - Alpha-Fold.

The more recent interest in AI, however, has turbocharged investment in AI and nowhere is this more prevalent than in the pharmaceutical sector. AI has countless applications for the sector including: in drug discovery, dosage form design, drug delivery and the optimisation of pharmacokinetic and pharmacodynamic processes, tests and studies. However, the regulation of this technology is fraught with complexity and controversy as regulatory frameworks struggle to adapt to rapid advances in AI technology. The regulatory challenge lies in creating frameworks that not only keep pace with innovation but also ensure safety, reliability, non-discrimination and adherence to ethical standards in the pharmaceutical sector. In this article we look at how the regulatory landscape in the EU and the UK is responding to these challenges and how the different approaches may affect development in the sector.

The Rapid growth of AI

AI is a leading technology today, encompassing machine learning, computer vision and natural language processing (NLP). It is expected to impact all industry sectors, potentially surpassing the computer and smartphone revolutions. The global AI software market is projected to grow by 54% annually, reaching $22.6 billion[1], while the AI healthcare market, valued at $11 billion in 2021, is estimated to reach $187 billion by 2030[2]. There has already been a substantial increase in the amount of capital invested in AI-driven pharma companies since 2015. During the last 9 years, the annual amount of investments in 800 companies has increased 27 times (to $59.3B in total as of December 2022)[3].

The rapid growth of AI, particularly since OpenAI launched ChatGPT in late 2022, has led to commercial AI applications reaching consumers with inadequate information about their training data and learning processes. This lack of transparency has raised alarms for regulators globally and these are particularly acute for healthcare applications. Concerns have been raised by researchers and human rights activists about the harm that AI is causing on the market and warning of dangerous realities such as AI-based medical devices that misdiagnose or disproportionately diagnose health problems in patients of a particular race. Despite these concerns, there are relevant sectors of society that believe that this technology should not be regulated fearing that regulations might slow down advances in AI and create barriers to entry for the creation of new companies. Additionally, they argue that balancing international norms and the sovereignty of each jurisdiction is very complex and would be an obstacle to the global development and launch of AI-based tools and technology. There is also a fear that that regulators may be unable to legislate in parallel with the evolution and speed of this technology.

Conversely, others argue for the possibility of self-regulation of the AI development industry thus avoiding the problems of competitiveness derived from state legislation but at the same time assuming ethical standards and criteria to avoid the risks of total lack of regulation.

Current Regulation

The issue of AI regulation, as well as the approach and nature that such regulation should have, is something that has been of particular concern to the European legislator. On 13 April 2024, the European Parliament officially approved the “AI Act” which aims to create uniform rules for AI systems in the EU. This legislation establishes four risks levels ranging from “minimal or no risk” to “unacceptable risk”. Notably, the AI Act exempts AI developed for scientific research from some of its provisions. The UK government on the other hand has chosen a less prescriptive approach to regulation, relying instead on a looser approach that relies on three core pillars, namely (i) investing in the long-term needs of the AI ecosystem (ii) ensuring that AI benefits all sectors and regions and (iii) governing AI effectively. It still remains to be seen on how these pillars will influence on AI development and investment in the UK vis-à-vis the EU. However, one interesting aspect of UK strategy is the focus on international technical standards as a mechanism for governing AI and to establish the UK as a ‘safe harbour’ for the development of AI technologies in the hope that it will encourage innovation.

In parallel with this new legislation, in December 2023, the European Medicines Agency (EMA) and the Network of Heads of Medicines Agencies (HMA) released an AI roadmap outlining a strategy up to 2028 to enhance AI’s role in the pharmaceutical sector. The roadmap aims to help the European Medicines Regulatory Network (EMRN) improve productivity, automate processes, boost data literacy and support robust decision-making for public and animal health. It focuses on supporting ongoing development and providing guidance for AI in medicine, establishing frameworks to use AI tools efficiently while ensuring data protection, advancing the network’s capacity to stay updated with AI developments and emphasising experimentation to accelerate learning and gain insights. The UK regulator, Medicines and Healthcare products Regulatory Agency (MHRA) also released a statement in April 2024 outlining its priorities for regulation of medical products. The focus of the MHRA appears to be towards a ‘pro-innovation’ approach to regulation of AI shaped by competing perspectives around its role as a regulator, a public service organization delivering time-critical decisions and one that makes evidence-based decisions that impact on public and patient safety. One the initiatives the MHRA has launched recently is a ‘AI-Airlock’ which is regulatory sandbox framework for trialling AI powered innovations using real world products and challenges.

Implications

In respect of the pharmaceutical and biotechnology fields, the European Federation of Pharmaceutical Industries Association (EFPIA) has outlined five key considerations regarding the introduction of the AI Act. Firstly, they support the exemption in the AI Act for AI used in scientific research, commending the Act’s support for innovation and scientific freedom. They note that most AI applications in drug research are unlikely to be classified as high-risk under the Act. Further, the EFPIA states that Europe's existing regulatory framework for drug development is sufficiently flexible to incorporate AI technologies and they endorse the EMA's proactive, risk-based approach to AI governance in medicines, as outlined in its roadmap. Finally, the EFPIA highlights the challenge of keeping regulations current with rapid AI advancements and calls for updates that consider AI's role, development stage, impact on drug benefit-risk assessments and the level of human oversight. Without prejudice to these statements, the EFPIA has also confirmed that it will look closely at the practical effects and shortcomings of the AI Act’s application to the field of drug research and development.

These overlap with the approach taken by the UK government which has identified five core principles for the regulatory use of AI:

• Safety, security and robustness
• Transparency and explainability
• Fairness
• Accountability and governance
• Contestability and redress

It is notable that both the European and the UK approaches are currently framed in terms of aspirational commitments rather than concrete policies for regulatory frameworks or governance. This is not particularly surprising at this early stage of the regulatory consulting process. However, as uptake of AI technology within the pharmaceutical industry increases, we expect that the inevitable trade-offs and competing stakeholder interests that ultimately underlie the transformation will means that the regulators will have to strike a delicate balance between ensuring patient safety and access as against allowing the technology to mature organically.

Future Direction

From our point of view, any regulation of AI should necessarily be developed within the framework of respect for a series of unavoidable principles, essentially centered on people and the benefit of society (as opposed to other interests, such as the purely economic, or indiscriminate technological progress). In healthcare and pharmaceuticals, AI must be safe, transparent, and ethical, ensuring patient data protection and bias-free outcomes. Strong intellectual property rules are also key to fostering biopharmaceutical innovation. While the European AI law addresses many of these concerns, regulators must stay adaptable to evolving global and technological changes to avoid future regulatory gaps. Similarly in the UK, where adaptability and a sector-led approach has to date taken precedence over legislation, the government will need to remain mindful of the risk of inconsistencies between sectors and gaps in regulation appearing.

For more information on AI in the Pharmaceutical Industry

Artificial intelligence workplan to guide use of AI in medicines regulation | European Medicines Agency (EMA) (europa.eu)

Further reading

How the regulatory landscape in the UK and EU is shaping the medical devices industry | Eversheds Sutherland (eversheds-sutherland.com)

Global AI Regulatory Update - September 2024 | Eversheds Sutherland (eversheds-sutherland.com)

Legal guide to navigating artificial intelligence - Legal issues around AI (eversheds-sutherland.com)

Global Health and Life Sciences newsletter - Edition 10 | Health and Life Sciences | Eversheds Sutherland (eversheds-sutherland.com)

Telemedicine: How laws and regulation are shaping the future of telemedicine in the UK | Eversheds Sutherland (eversheds-sutherland.com)

[1] Forecast growth of the artificial intelligence (AI) software market worldwide from 2019 to 2025, Statista, Inc, April 2019.

[2]BERGER E., SANDIG R., and GEORGE, KC., "How to Successfully Scale Generative AI in Pharma," in Bain & Company, February 12, 2024.

[3] Research carried out by Deep Pharma Intelligence in 2023