LSHC Horizons Brochure 2024 - Flipbook - Page 16
Hogan Lovells | 2024 Life Sciences and Health Care Horizons | Digital Health and AI
16
How the AI Act and the EHDS will influence research
and development with AI in health care
Health care companies are facing and will
continue to face challenges when developing AI
systems that use health data. Such challenges
can result from the interaction of the AI Act
with other EU regulations like the Medical
Device Regulation (MDR), the EU General
Data Protection Regulation (GDPR) and the
upcoming European Health Data
Space (EHDS).
In the evolving landscape of health care and AI,
the synergy between the AI Act and the EHDS
plays a pivotal role. This interplay is especially
significant for AI developers, who could benefit
immensely from the vast reserves of health data
available in Europe.
In the context of the EHDS, a key challenge lies
in the potential limitations on the availability
of electronic health data (eHD), particularly
regarding their utilization as training datasets
for AI systems. Should the input data be limited
or not fully representative, it could lead to
biased and imbalanced outcomes. Such skewed
data can significantly impact the effectiveness
and development trajectory of AI systems.
This risk shows the importance of diverse
and comprehensive datasets to ensure that AI
algorithms in health care are well-informed
and unbiased, ultimately contributing to
robustness and reliability.
Patrice Navarro
Partner
Paris
The concurrent application of the AI Act and
the EHDS presents a complex regulatory
landscape. AI systems that employ eHD for
training are subject to both regulations (as
per Art. 1 of the EHDS draft). The European
Parliament’s proposal for an opt-out
mechanism for general eHD and consent-based
access to particularly sensitive data, such as
genetic information, introduces challenges in
data accessibility. This could lead to a scarcity
of certain types of health data or health data
from a certain peer group with the risk of
imbalanced and biased AI outcomes due to
non-representative datasets. The AI Act's
provision to correct bias by permitting, under
specific conditions, the use of sensitive data
as per Art. 9(1) GDPR, may not adequately
address these imbalances in the health care
sector. The success of the AI Act in promoting
non-discriminatory AI systems in the EU
hinges on the EHDS’s ability to facilitate
equitable access to health data.
In conclusion, while the AI Act aims to
ensure that AI systems used in the EU are
fair and non-discriminatory, the success of
this objective in the health care sector largely
depends on the final formulation of the
EHDS. We are optimistic that the EHDS, in
its finalized version, will effectively support
this goal by facilitating access to a diverse and
balanced range of health data.
Arne Thiermann
LL.M. (LSE)
Partner
Hamburg
Dr. Karolin Hiller
Counsel
Munich, Berlin
