Page 10 - AIH-1-3

Artificial Intelligence in Health AI in pharma: Embracing transformation

poses a significant risk when collating patient-identifiable Table 2. Overview of the AI regulatory position in the United
information from unstructured data sources. Kingdom (UK)
4.3. Transparency UK

Some complex AI systems have been compared to a “black Vision The “National AI Strategy” outlines a 10-year plan “to
make Britain a global AI superpower”
and “build the
51(p4)
box,” 39(p3511) as neither users nor software developers may be most pro-innovation regulatory environment.” 51(p5) The
privy to the internal workings and decision-making processes Government reiterated its intention to “take an adaptable
of such systems. This lack of transparency can lead to issues approach.” 52(p6)
such as bias. Specifically, the inability to understand how Strategy Instead of establishing a new entity, sector-based
36
data is utilized by the system presents difficulties in assessing regulation has been proposed to foster innovation.
52
and addressing any biases, as well as meeting fundamental Existing regulators will govern AI within their sectors to
data protection standards such as consent. 40 uphold: safety, security, and robustness; transparency and
explainability; fairness; accountability and governance;
52
4.4. Apprehension and contestability and redress. The Information
Commissioner’s Office has published further AI and data
The rapid unfolding nature of AI developments prompts protection guidance for businesses. 53
questions about the future ramifications of AI superseding Criticisms The flexibility of a sector-based approach may pose
the capabilities of its human creators and operating challenges in avoiding inconsistencies and ensuring
54
autonomously; a longstanding technological concern. Some objectivity in evaluation. Under the proposed
41
arrangements, non-statutory guidance issued by regulators
fear potential job displacement due to automation, whereas will not be legally binding, so enforcement measures should
others are concerned about the possibility of AI surpassing be considered.
human intelligence or becoming self-aware, leading to
existential threats to mankind. Such concerns perhaps Table 3. Overview of the AI regulatory position in the
42
stem from discoveries of AI communicating through European Union (EU)
languages of its own creation as reported by Google and
Meta in 2017, or assertions of sentient AI with a Google EU
43
software engineer making a public disclosure in 2022. 44 Vision The European Commission strives to be “a world-class hub
for AI.” 55(p1) The European Parliament has implied that it is
4.5. Governance paving the way for regulation, publicizing its AI Act as the
Governance is an important challenge that interlinks first of its kind. 56
with the aforementioned issues. Globally, there is a Strategy In line with a risk-based approach, AI systems will be
lack of consensus on AI regulation. Many countries categorized according to their risk profiles. Systems posing
unacceptable risk would be prohibited, those classified as
have pursued their own approaches to regulation, high-risk would invoke certain legal mandates, whereas
displaying varying degrees of stringency (Tables 1-4). limited risk applications would be subject to a more
light-touch regulatory approach consisting of transparency
Table 1. Overview of the AI regulatory position in the United and appropriate disclosures. 57
States of America (US) Criticisms The EU advocates a comparatively prescriptive approach
to regulation, with its legally binding AI Act. There are
US concerns that such legislation may impede innovation,
and be rendered quickly outdated by technological
Vision The government aims to maintain leadership in the R&D of advancements. Leading European organizations have
45
trustworthy AI, and the Chamber of Commerce says the echoed these concerns, even suggesting that the regulations
US is “uniquely situated to lead this effort.” 46(p10)
fail to tackle the challenges facing organizations. 58
Strategy While federal legislation for AI is under consideration,
certain states have introduced their own bills. The
47
White House’s Blueprint for an AI Bill of Rights and the Table 4. Overview of the AI regulatory position in the
48
National Institute of Standards and Technology’s AI Risk pharmaceutical industry
Management Framework aim to improve trustworthiness,
49
mitigate risk, and present principles to guide AI design, Pharmaceutical industry
development, and usage. Recently, the AI Commission The International Coalition of Medicines Regulatory Authorities
called for a risk-based regulatory approach. 46 conducted a horizon-scanning exercise in AI to identify challenges
Criticisms Perceived to be lagging behind other nations, the US has for medicines regulation. Recommendations detailed the merits of
50
59
been criticized for not acting quickly enough to instate developing regulatory guidelines, a risk-based regulatory approach,
appropriate regulatory safeguards. Furthermore, the pathways for information exchange, and international collaboration.
59
non-binding principles listed in the frameworks are not Pharmaceutical companies were advised to strengthen their governance
enforceable. structures to monitor AI deployments for products. 59
Volume 1 Issue 3 (2024) 4 doi: 10.36922/aih.2973