UK Trends and Developments Contributed by: Amélie Chollet, Hannah Curtis and David Dennis, CMS
Product liability: what happens when algorithms make mistakes? The introduction of AI into healthcare products creates novel liability questions that traditional frameworks struggle to address. From the outset, where an AI system is impli- cated in a product liability issue, assessing causation and determining who may be the responsible party can become significantly more complex: for example, when an AI-pow- ered diagnostic medical device misses a critical finding that leads to a patient injury, determining liability between the healthcare provider, device manufacturer, algorithm developer and training data provider requires new legal frameworks and approaches. The use of AI also presents another challenge in the context of establishing the standard of care. When the care delivered involves AI-powered technology that is constantly evolving through machine-learning (ML) capabilities, how does one determine what is the appropriate standard of care? As these systems continue to develop and change, the benchmark against which rea- sonable care is measured becomes a moving target. Questions of transparency and explainability may further complicate liability determinations. Courts and regulators will need to establish what level of transparency and explainability is required to demonstrate that reasonable care was taken during development. Additionally, managing discovery and disclosure of algo- rithmic processes during litigation introduces procedural challenges not present in traditional product liability cases.
Regulatory compliance: navigating an evolving patchwork Life sciences companies now face a complex regulatory ecosystem that spans both estab- lished frameworks and emerging AI-specific regimes. This landscape includes the MHRA’s evolving work on software as a medical device guidance, the EU’s new AI Act with its risk-based approach, and the FDA’s proposed framework for AI/ML-enabled medical devices. Simultane- ously, existing regulations such as the Medical Device Regulation (MDR) and the In Vitro Diag- nostic Medical Devices Regulation (IVDR) are being interpreted and applied in AI contexts, creating additional layers of complexity. Perhaps the most vexing challenge for multi- national life sciences companies is navigating this global patchwork of AI regulation, where fundamental inconsistencies can result in sig- nificant – and costly – compliance hurdles. Dif- ferent jurisdictions employ varying terminology and definitions – the EU has established spe- cific high-risk categories, while the UK and USA take different approaches to classification. The very definition of what constitutes “AI” versus conventional software varies across regulatory regimes, creating fundamental challenges in determining which rules apply to specific prod- ucts. This can also complicate the design and implementation of internal processes across dif- ferent markets and that will be subject to regula- tory audits. Requirements for human oversight and interven- tion also differ substantially between jurisdic- tions, forcing companies to implement different governance models depending on the market. Beyond these definitional differences, proce- dural requirements diverge significantly across borders. Documentation standards, evidence
141 CHAMBERS.COM
Powered by FlippingBook