An AI A Day Keeps The Doctor Away?: Regulating Artificial Intelligence In Healthcare – Healthcare

Artificial Intelligence (“AI”) guarantees to remodel
many facets of on a regular basis life for Canadians. AI instruments are predicted
to dramatically enhance the supply of heath care by bettering
the standard, security, and effectivity of diagnostic instruments, therapy
choices, and care. Although AI improvements are, in lots of circumstances,
nonetheless years away from common deployment into the Canadian well being
care ecosystem, AI is already utilized in some circumstances to learn
medical photographs, permitting machine studying to help diagnosticians
of their resolution-making.

Like many different jurisdictions, Canada’s well being governance
techniques at the moment lack the suitable authorized and regulatory
mechanisms to successfully cope with the challenges that AI poses.
There is at the moment uncertainty with respect to key points equivalent to
the associated authorized necessities for well being privateness, medical gadget
regulation and legal responsibility for AI-related harms. In Canada,
regulation of AI in well being care includes the extra problem
of navigating constitutionally fragmented jurisdiction over well being
care, which ends up in layers of governance and the necessity to
coordinate a number of completely different actors.

This weblog put up highlights among the authorized challenges and
points that should be addressed to ensure that Canada to have a
sturdy and nicely-regulated governance construction for the usage of AI in
well being care, together with:

Coordination of federal and provincial authority;

Privacy and oversight with respect to the usage of AI in

Promotion of Equity by means of AI; and

Liability for AI-related harms.

Coordination of Federal and Provincial Authority

Canada’s federal system and constitutional division of
powers pose distinctive challenges for the regulation of AI in well being
care.1 Under the Constitution, well being care is beneath
provincial jurisdiction. Although comparable, every province has its
personal set of regulatory frameworks addressing the security and high quality
of well being care, well being data privateness, knowledgeable consent, human
rights and non-discrimination, and licensing of well being care
professionals. With respect to the adoption of AI, provincial
laws and regulation would be the major authorized construction that
governs the top customers of AI know-how and its utility to

However, regardless of well being care being primarily a provincial
concern, the federal authorities performs a big function,
notably by means of its spending powers beneath the Canada
Health Act,2 and its obligations for
Indigenous Peoples, federal prisoners, and the navy. The
federal authorities can also be a big participant within the regulation
of medicine and medical units. 

Health Canada is the important thing regulatory authority on the federal
stage that controls which medical units can be found on the market
and could also be included within the public insurance coverage of the provinces
and territories. Health Canada’s major mode of regulation is
by means of the licensing course of relevant to all medical units.
This licensing course of requires producers to categorise their
units in keeping with danger (e.g., invasiveness, danger of faulty
analysis, and supposed medical function) beneath the Medical
Devices Regulations,3 and procure approval from
Health Canada. If a tool is licensed, the Medical Device
Directorate continues to observe the security and efficacy of the

A important problem for the licensing and regulation of
AI-technology is the applying of machine studying, which is
sometimes called “black-field” resolution making, as a result of
the related algorithms are sometimes proprietary and commercially
delicate and choices and impacts of the algorithms can’t be
totally defined. A query that’s being requested by regulators
all over the world is “how can a regulator confirm and validate
machine studying algorithms to make sure that they do what they are saying
nicely and safely?”5 Another query is: what function
ought to machine studying and automatic resolution making have in
well being care?

Other key actors within the regulatory framework of Canadian well being
care are the skilled our bodies that present oversight and
self-regulation. Ensuring coordination between these regulatory
our bodies, the provincial and federal legislatures, and Health Canada
to reduce or eradicate regulatory blind-spots will probably be a problem
that have to be overcome to make sure good governance of AI in well being

In April 2021, the European Commission launched a 108-web page
proposal to control AI. Although the European Union has but to
attain consensus on the ultimate textual content of the laws, the proposal
has obtained important curiosity, and the query of how the
European mannequin may inform the event of AI governance in
Canada is being thought of by thought leaders in Canada.

Privacy and Oversight

Even although the European Union has not but applied an
overarching framework for AI regulation, the principles in its General
Data Protection Regulation (“GDPR”) present important
steerage for the European medical group with respect to
regulation of medical AI.7 For instance, beneath the GDPR,
any controller of an AI-system based mostly solely on automated processing
should present the topic with details about the existence of
the automated resolution-making, significant details about the
logic concerned, and the importance and penalties of such
processing.8 The GDPR additionally offers a sturdy regulatory
framework that governs the info privateness of residents whose knowledge could
be utilized in machine studying algorithms. In Europe, the GDPR additionally
requires that medical AI have human oversight.9 Further,
the coaching knowledge for the AI have to be checked for bias and the
ongoing operation of AI have to be continuously monitored for the
incidence of bias to make sure that use of AI doesn’t
unintentionally lead to discrimination.10 

Recent modifications to Canada’s provincial privateness panorama
means that Canada won’t solely observe the European instance, however
additionally search to implement sturdy privateness rights in its personal method. For
occasion, on September 22, 2021, the province of Quebec’s
landmark laws, the Act to Modernize Legislative
Provisions respecting the Protection of Personal Information
(“Bill 64”), obtained Royal Assent. Bill 64 will impose a
obligation to tell with respect to technological instruments that allow the
identification, location or profiling of a person with the intention to
gather private data from a person. From September 22,
2023, organizations may even be required to tell the person
when a choice is made based mostly solely on automated processing of his
or her private data, no later than the time the
group informs the person of that call. Organizations
shall additionally give the person the chance to make
representations to a member of their employees who is able to
evaluation the choice. For extra details about Bill 64, seek the advice of
our weblog sequence right here.

Similarly, the federal authorities’s Directive on Automated
Decision Making (“Canada ADM Directive”),11
signifies that Canadian regulatory frameworks may even possible
require that any well being-centered AI know-how present for human
intervention within the resolution-making course of, and be certain that all
knowledge be examined for bias and non-discrimination.12 The
Canada ADM Directive is a danger-based mostly governance mannequin that
establishes 4 ranges of danger, judged by the impression of the
automated resolution. Certain danger-mitigating necessities are then
established for every impression stage, together with: discover earlier than
automated resolution making choices and explanations after
automated resolution making choices; peer evaluation; worker
coaching; and human intervention.

Promotion of Equity

Arguably essentially the most important concern related to the usage of
AI and automatic resolution making in well being care is their potential
to amplify bias and discrimination. Canada’s well being care system
already grapples with the issues related to inequities in
well being care – from differential useful resource allocation between
communities, to differential therapy of people based mostly upon
their gender or race.

The present authorized framework for funding well being care in Canada
(i.e., the Canada Health Act) solely protects common
protection for “medically mandatory” hospital and doctor
providers.13 Due to the present novelty of AI-assisted
medical providers, it’s unlikely that many AI-assisted medical
providers would at the moment be thought of medically mandatory.
Therefore, solely sufferers who’ve the means to afford add-on charges
or non-public or boutique well being care would acquire entry to this
refined know-how. Further, if solely bigger medical facilities
have the infrastructure and entry to such laptop science
applications essential to develop AI-assisted medical applications, entry
to AI-technology could also be restricted, even when value will not be a barrier.
Therefore, regulating AI by means of the suitable authorized frameworks
to make sure that it’s developed and deployed in an accessible method
will probably be an vital matter for legislatures to contemplate and

In addition to potential inequities of entry, there are two
primary sources of concern referring to discrimination in AI techniques:
(1) bias within the knowledge used to coach the system; and (2) bias within the

If the info used to coach the AI system is flawed or incomplete,
for instance by failing to incorporate adequate knowledge from a sure
inhabitants, the AI system could also be ineffective or harmful for
sufferers of the underrepresented inhabitants. For instance,
AI-assisted most cancers screening instruments which might be educated totally on
photographs of sunshine-skinned sufferers usually tend to misdiagnose
most cancers lesions in sufferers with pores and skin of color.14 Bias
within the AI-training knowledge, whereas simpler to determine, poses severe
questions referring to entry to knowledge, knowledge switch, and consent.
The significance of coaching AI techniques with knowledge from numerous
populations should be balanced with legal guidelines referring to knowledge
assortment, use, switch and storage throughout a number of
jurisdictions, in order that numerous sufferers residing in areas with much less
numerous populations aren’t prone to being harmed by therapies
based mostly on an absence of numerous knowledge.

Bias within the algorithm of an AI system could also be not possible detect,
notably the place machine studying strategies are employed. When
the choice making of the AI is a “black field”, attributable to
the opacity of how the AI is figuring out patterns (and probably
to modifications within the algorithm over time because the machine learns), it
is usually a problem to make sure that discrimination will not be occurring.
To fight this, a number of jurisdictions are contemplating express
legislative commitments that guarantee AI techniques are compliant with
anti-discrimination and human rights laws.15
Paired with sturdy monitoring necessities, most of these
provisions would supply higher authorized certainty, accountability,
and public confidence in AI-assisted well being care.

Liability for AI-Related Harms

Another appreciable hurdle within the adoption of AI applied sciences
in well being care, notably within the medical group, is the
continued uncertainty concerning the potential legal responsibility hooked up to
the usage of AI. Who do you sue when AI goes flawed?

Although AI has been used for numerous functions over the previous
few years, it stays unclear the place legal responsibility ought to fall when an
AI system fails. In 2020, the query of who (if anybody) is liable
when an AI-powered buying and selling funding system causes substantial
losses for an investor, was earlier than the English courts for the primary
time.16 Unfortunately for the event of tort legislation in
this vital space, the events reached an out of court docket
settlement, leaving the query to be answered one other

In Canada, medical harms could also be handled beneath the legislation of
negligence.18 How AI-technology modifications the usual of
care anticipated of a medical practitioner and, particularly, the
acceptable stage of resolution-making delegation to the AI system,
are questions that have to be thought of.19 Further, the
courts should question whether or not an AI firm has any legal responsibility to a
affected person that’s misdiagnosed. Could an AI firm contract out of
its legal responsibility to a hospital, if hurt outcomes from the usage of its
know-how? Is an AI firm solely liable if bias is discovered within the
knowledge or algorithm? What kind of consent is required from sufferers
earlier than AI applied sciences are employed?

Another space the place AI could trigger important hurt to Canadians is
by means of breaches of personal well being care knowledge, which may hurt the
public’s total confidence within the well being care system. In the
current Supreme Court of Canada resolution, Reference re Genetic
Non-Discrimination Act (2020 SCC 17), the bulk held that
the federal authorities had the facility to make guidelines combating
genetic discrimination and defending well being by means of its
jurisdiction over legal legislation:

Many of this Court’s choices illustrate how the legal
legislation function check operates. A legislation directed at defending a public
curiosity like public security, well being or morality will often be a
response to one thing that Parliament sees as posing a menace to
that public curiosity. For instance, prohibitions geared toward combatting
tobacco consumption and defending the general public from adulterated
meals and medicines have been upheld as a result of they shield public well being from
threats to it…

Parliament took motion in response to its concern that
people’ vulnerability to genetic discrimination posed a
menace of hurt to a number of public pursuits historically protected
by the legal legislation. Parliament enacted laws that, in pith
and substance, protects people’ management over their
detailed private data disclosed by genetic checks within the
areas of contracting and the supply of products and providers in
order to handle Canadian’s fears that their genetic check
outcomes will probably be used towards them and to forestall discrimination
based mostly on that data. It did so to safeguard autonomy, privateness
and equality, together with public well being. The challenged provisions
fall inside Parliament’s legal legislation energy as a result of they
encompass prohibitions accompanied by penalties, backed by a
legal legislation function.20

This case demonstrates that the federal authorities’s
regulatory energy will not be restricted to well being care spending. However,
it’s unclear how successfully legal legislation can be utilized to manipulate AI
and, as AI know-how turns into extra widespread in well being care, the
legislatures and the courts should fastidiously take into account how the
present non-public and legal legislation frameworks may be tailored to deal
with attributing and apportioning legal responsibility arising from AI


AI poses each dangers and alternatives within the well being care house.
AI techniques purpose to democratize well being and supply superior affected person
care. However, regulators should cope with the problem of
making certain that AI know-how does what it’s supposed to do, does it
nicely, and that there stays authorized accountability for any harms
brought about.


1 Colleen M. Flood and Catherine Régis,
Régis, Catherine and Flood, Colleen M., AI and Health Law
(February 1, 2021). in Florian Martin-Bariteau & Teresa Scassa,
eds., Artificial Intelligence and the Law in Canada (Toronto:
LexisNexis Canada, 2021), Available at SSRN:

2 Canada Health Act, R.S.C., 1985, c.

3 Medial Devices Regulations, SOR


5 W. Nicholson Price II, “Artificial Intelligence in
Health Care: Applications and Legal Issues” (2017) The SciTech
Lawyer 14:1; David Schneeberger et al., (2020) The European Legal
Framework for Medical AI. In: Holzinger A., Kieseberg P., Tjoa A.,
Weippl E. (eds) Machine Learning and Knowledge Extraction. CD-MAKE
2020. Lecture Notes in Computer Science, vol 12279. Springer, Cham.

6 Law Commission of Ontario, “Comparing European and
Canadian AI Regulation” (November 2021).

7 Regulation (EU) 2016/679 of the European Parliament and
of the Council of 27 April 2016 on the safety of pure
individuals with regard to the processing of non-public knowledge and on the
free motion of such knowledge, and repealing Directive 95/46/EC
(General Data Protection Regulation) [GDPR]

8 GDPR, ibid, Arts 13, 14.

9 GDPR, ibid, Art 22. Scheenberge, supra
notice 5, 211.

10 Schneeberger, supra notice 5, at

11 Canada ADM Directive.

12 Federal Government’s Directive on Automated
Decision-Making: Considerations and Recommendations

13 Canada Health Act, R.S.C., 1985, c. C-6., s.
2 sub nom “hospital providers” and “doctor

14 Adamson AS, Smith A. Machine Learning and Health Care
Disparities in Dermatology. JAMA Dermatol. 2018 Nov
1;154(11):1247-1248. doi: 10.1001/jamadermatol.2018.2348. PMID:

15 Law Commission of Ontario, “Regulating AI:
Critical Issues and Choices” (April 2021) at 38-39.

16 Minesh Tanna “AI-powered investments: Who (if
anybody) is liable when it goes flawed? Tyndaris v VWM”
(November 2019).

17 Jeremy Kahn, “Why accomplish that few enterprise see
monetary features from utilizing A.I?” (October 20, 2020).

18 In Quebec, civil legislation legal responsibility ideas would

19 Mélanie B. Forcier, et al., “Liability
points for the usage of synthetic intelligence in well being care in
Canada: AI and medical resolution-making” (July 2020) Dalhousie
Medical Journal 46(2). DOI:10.15273/dmj.Vol46No2.10140

20 Reference re Genetic
Non-Discrimination Act, 2020 SCC 17 at
paras. 73, 103

To view the unique article click on right here

The content material of this text is meant to supply a common
information to the subject material. Specialist recommendation must be sought
about your particular circumstances.

Recommended For You