174
Chapter 11: Artificial Intelligence
AI,’61 for example, includes both algorithmic
as well as data aspects: non-discrimination,
diversity, bias, privacy and data governance,
societal and environmental wellbeing (e.g.,
energy needed to train AI and its impact on
global warming), human agency and oversight,
transparency, accountability (related to logging),
technical robustness, and safety. Most of these
aspects are already covered by existing regula-
tions, such as the medical device, data protection,
product liability, and CO2 emission regulations,
although they are not as specific to AI as digital
rights advocates and the European Parliament
would like. The European Commission intends
to publish a legislative proposal on 21 April
2021 to address the ethical aspects of AI. It will
likely cover several of these ethical dimensions
and apply on a horizontal level, i.e., potentially
include machine learning medical devices. The
EU is also assessing whether and how to address
the liability and intellectual property of AI.
Chinese AI Legislation
The Organization Center for Medical Device
Evaluation (CMDE), a division of the Chinese
regulatory authority, the National Medical Prod-
ucts Administration (NMPA), issued compre-
hensive requirements62 encompassing scrutiny
of machine learning devices across their entire
device lifecycle. The National Institute for Food
and Drug control (NIFDC), another division of
NMPA, supplements these requirements with a
growing body of standards, e.g., to characterize
the data sets used for the training or evalua-
tion of AI (see Figure 11-11). Manufacturers
are well-advised to take these recommended
standards into account early in the development
project when drafting the clinical evaluation
and clinical development plan so that there are
no surprises when the regulatory submission for
China is prepared.
Through standardization, the Artificial Intel-
ligence Medical Device Innovation and Coop-
Figure 11-10. Geographic Distribution of Issuers of Ethical AI Guidelines by Number of
Documents Released
Source: Jobin A. Artificial Intelligence: The global landscape of ethics guidelines. June 2019. https://www.
researchgate.net/publication/334082218_Artificial_Intelligence_the_global_landscape_of_ethics_guidelines.
Accessed 16 February 2021.
1 guideline
Between 2-4 guidelines
No guidelines
Between 5-14 guidelines
15 or more guidelines
Chapter 11: Artificial Intelligence
AI,’61 for example, includes both algorithmic
as well as data aspects: non-discrimination,
diversity, bias, privacy and data governance,
societal and environmental wellbeing (e.g.,
energy needed to train AI and its impact on
global warming), human agency and oversight,
transparency, accountability (related to logging),
technical robustness, and safety. Most of these
aspects are already covered by existing regula-
tions, such as the medical device, data protection,
product liability, and CO2 emission regulations,
although they are not as specific to AI as digital
rights advocates and the European Parliament
would like. The European Commission intends
to publish a legislative proposal on 21 April
2021 to address the ethical aspects of AI. It will
likely cover several of these ethical dimensions
and apply on a horizontal level, i.e., potentially
include machine learning medical devices. The
EU is also assessing whether and how to address
the liability and intellectual property of AI.
Chinese AI Legislation
The Organization Center for Medical Device
Evaluation (CMDE), a division of the Chinese
regulatory authority, the National Medical Prod-
ucts Administration (NMPA), issued compre-
hensive requirements62 encompassing scrutiny
of machine learning devices across their entire
device lifecycle. The National Institute for Food
and Drug control (NIFDC), another division of
NMPA, supplements these requirements with a
growing body of standards, e.g., to characterize
the data sets used for the training or evalua-
tion of AI (see Figure 11-11). Manufacturers
are well-advised to take these recommended
standards into account early in the development
project when drafting the clinical evaluation
and clinical development plan so that there are
no surprises when the regulatory submission for
China is prepared.
Through standardization, the Artificial Intel-
ligence Medical Device Innovation and Coop-
Figure 11-10. Geographic Distribution of Issuers of Ethical AI Guidelines by Number of
Documents Released
Source: Jobin A. Artificial Intelligence: The global landscape of ethics guidelines. June 2019. https://www.
researchgate.net/publication/334082218_Artificial_Intelligence_the_global_landscape_of_ethics_guidelines.
Accessed 16 February 2021.
1 guideline
Between 2-4 guidelines
No guidelines
Between 5-14 guidelines
15 or more guidelines