Access to this full-text is provided by Taylor & Francis.
Content available from Applied Artificial Intelligence
This content is subject to copyright. Terms and conditions apply.
AI Ethics: Integrating Transparency, Fairness, and Privacy in
AI Development
Petar Radanliev
Department of Computer Science, University of Oxford, Oxford, UK
ABSTRACT
The expansion of Articial Intelligence in sectors such as health-
care, nance, and communication has raised critical ethical
concerns surrounding transparency, fairness, and privacy.
Addressing these issues is essential for the responsible devel-
opment and deployment of AI systems. This research estab-
lishes a comprehensive ethical framework that mitigates
biases and promotes accountability in AI technologies.
A comparative analysis of international AI policy frameworks
from regions including the European Union, United States, and
China is conducted using analytical tools such as Venn diagrams
and Cartesian graphs. These tools allow for a visual and sys-
tematic evaluation of the ethical principles guiding AI develop-
ment across dierent jurisdictions. The results reveal signicant
variations in how global regions prioritize transparency, fairness,
and privacy, with challenges in creating a unied ethical stan-
dard. To address these challenges, we propose technical strate-
gies, including fairness-aware algorithms, routine audits, and
the establishment of diverse development teams to ensure
ethical AI practices. This paper provides actionable recommen-
dations for integrating ethical oversight into the AI lifecycle,
advocating for the creation of AI systems that are both techni-
cally sophisticated and aligned with societal values. The ndings
underscore the necessity of global collaboration in fostering
ethical AI development.
ARTICLE HISTORY
Received 11 August 2024
Revised 5 September 2024
Accepted 2 February 2025
Introduction
In recent years, substantial advancements in AI ethics have emerged, with
significant contributions addressing transparency, fairness, and privacy in AI
development. Recent research studies (Bender et al. 2021) highlight the dan-
gers of bias in large language models, raising concerns over the perpetuation of
societal inequalities within AI systems. Similarly, Bommasani et al. (2023)
critically evaluate compliance of foundation models with the draft EU AI Act,
reflecting broader concerns over the accountability of AI systems at
a foundational level. Moreover, Aldoseri, Al-Khalifa, and Hamouda (2023)
CONTACT Petar Radanliev petar.radanliev@cs.ox.ac.uk Department of Computer Science, University of
Oxford, Oxford, UK
APPLIED ARTIFICIAL INTELLIGENCE
2025, VOL. 39, NO. 1, e2463722 (41 pages)
https://doi.org/10.1080/08839514.2025.2463722
© 2025 The Author(s). Published with license by Taylor & Francis Group, LLC.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/
licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s)
or with their consent.
propose new data strategies for AI development, focusing on the integration of
ethical principles across diverse datasets to mitigate bias. These works, along-
side key regulatory frameworks from bodies such as the European Union
(European Parliament 2023) and NIST (2024b), reinforce the imperative of
developing AI systems that are not only transparent and fair but also respect
data privacy and societal norms. By situating this study within the context of
these contemporary advancements, the aim is to build upon these discussions
by offering a comparative analysis of global AI policy frameworks, extending
the dialog on how ethical AI can be systematically developed and maintained
across diverse geopolitical contexts.
While this study references key AI policy frameworks from the European
Union, the United States, and China, the focus of this study is a more granular
examination of the challenges in comparing such diverse frameworks is
crucial. AI ethics policies are deeply influenced by cultural, socio-political,
and economic contexts, making cross-regional comparisons inherently com-
plex. For instance, the European Union’s AI Act places significant emphasis
on safeguarding individual rights, prioritizing transparency and human over-
sight (European Parliament 2023), reflecting the EU’s regulatory ethos aimed
at protecting citizens from the potential harms of AI. In contrast, the United
States’ AI governance is more decentralized, with a focus on promoting
innovation and maintaining global technological leadership, as seen in the
National Institute of Standards and Technology (NIST) AI Risk Management
Framework (2023a), which advocates for flexible, non-prescriptive guidelines
that encourage industry-led solutions.
China’s AI policy framework, meanwhile, is characterized by its focus on
state security, social harmony, and the integration of AI into national eco-
nomic strategies (Roberts et al. 2021). This framework aligns with China’s
broader governmental control over technology, where the state plays a central
role in guiding AI development. These diverging priorities highlight the
inherent challenges in creating universal standards for AI ethics. The task of
comparing these frameworks, therefore, requires consideration of their dis-
tinct legal, cultural, and economic motivations, as well as the varying levels of
public trust in AI technologies across these regions.
This study addresses these complexities by identifying common ethical
principles such as fairness, transparency, and privacy, and by analyzing how
each region interprets and prioritizes these principles. By employing compara-
tive tools such as Venn diagrams and Cartesian graphs, the article visually and
analytically demonstrates the differences, but also the common points in these
frameworks. The aim of this study was not to look only for the differences, but
to find a solution for global AI governance and to promote the potential for
harmonization across jurisdictions.
This paper explores the pressing need for ethical considerations in the
rapidly evolving domain of Artificial Intelligence (AI) (Meissner 2020). This
e2463722-2 P. RADANLIEV
technology has significantly impacted various sectors, including healthcare,
finance, and communication. This study aims to establish a robust ethical
framework for AI development by addressing complex issues such as data
privacy, algorithmic transparency, and fairness. Our objectives include analyz-
ing fundamental ethical principles, comparing international AI policy frame-
works (Helbing et al. 2018), proposing strategies for bias mitigation, and
contributing to academic and practical discussions in AI ethics. This paper
is structured to systematically dissect these topics, providing an in-depth
exploration of AI ethics and its implications for future AI development and
governance (de Fine Licht and de Fine Licht 2020).
The Imperative of Ethical Considerations in AI
The advent of Artificial Intelligence (AI) has inaugurated a new epoch in
technological evolution, profoundly influencing diverse sectors, including
healthcare, finance, transportation, and communication (Hosny et al. 2018;
NIST 2023b; Yu, Beam, and Kohane 2018). This unprecedented integration of
AI into the societal fabric necessitates the urgent formulation of robust ethical
frameworks. These frameworks must address the complexities inherent in AI
technologies, such as data privacy, algorithmic opacity, equity in decision-
making, and broader societal impacts.
Ethical considerations in AI transcend academic discourse, bearing signifi-
cant real-world repercussions. Paramount among these are issues related to
data privacy and the need for informed consent, where personal information
often powers AI algorithms. Equally critical is the transparency and explic-
ability of these algorithms, which are essential for sustaining public trust,
especially in high-stakes scenarios like legal adjudication or medical diagnos-
tics. Moreover, the challenge of ensuring equity and circumventing ingrained
biases in AI systems is a pivotal ethical imperative, given these systems’
propensity to mirror and perpetuate existing societal disparities.
The need for ethical AI is driven by the imperatives of harm prevention and
justice but also by the strategic objective of nurturing sustainable, socially
beneficial, and universally accepted innovation.
Aims and Objectives of the Study
The primary goals of this academic study are threefold. First, it seeks to
explore the ethical considerations inherent in the development of AI. This
involves thoroughly examining the fundamental ethical principles of transpar-
ency, equity, and privacy within AI systems and understanding how they relate
to each other and their significance in isolation. Second, the research aims to
critically analyze various global AI policy frameworks, focusing on those from
the EU, the US, and China. The goal is to discern their similarities and
APPLIED ARTIFICIAL INTELLIGENCE e2463722-3
differences and what they mean for international AI governance. Third, the
paper intends to provide a synthesis of approaches and practices to recognize
and mitigate bias in AI systems, ensuring their fairness and dependability. This
study aims to contribute to the ongoing academic and practitioner dialog on
AI ethics by offering relevant insights and recommendations to both groups.
The overarching goal is to promote a more ethical and responsible path for AI
development.
Structure and Content of the Paper
The paper has been methodically structured to explore AI ethics across various
dimensions systematically. Section 2, expands into the foundational ethical
principles in AI: transparency, equity, and privacy. The section uses a Venn
diagram to demonstrate the interaction between these principles, emphasizing
their interconnectedness and how they relate to AI.
Moving on to section 3, the focus is on integrating global AI policy frame-
works within AI development and deployment processes. The section presents
a flowchart outlining the critical stages in AI projects and the influence of
various international frameworks. This section examines the role of policies in
ensuring responsible AI development and the importance of incorporating
international frameworks to achieve this goal.
Section 4, provides a comparative analysis of AI Ethics Policy Frameworks
from different nations, using a Cartesian graph for evaluation based on
transparency, accountability, equity, and privacy. This seciton highlights the
varying approaches different countries take to AI ethics policies and how they
compare.
Section 5, proposes a range of strategies for addressing and reducing bias in
AI systems. It emphasizes the significance of data diversity, rigorous audits,
ethical training, and algorithmic clarity in reducing bias in AI systems.
Finally, the paper concludes with section 6, which summarizes the key
findings, discusses their implications for the future trajectory of AI develop-
ment, and suggests avenues for further scholarly inquiry. The paper provides
a comprehensive, multifaceted examination of AI ethics through this struc-
tured approach, contributing substantive insights to the ongoing scholarly
dialog in this critically pivotal domain.
Ethical Considerations in AI Development
Ethical considerations are of the utmost importance in the development of AI
to ensure that these systems are safe, fair, and transparent. A Venn diagram
illustrates the interplay between three key aspects: transparency, fairness, and
privacy.
e2463722-4 P. RADANLIEV
Transparency, Explainability, and Clarity refer to making AI systems clear
and understandable to users. Fairnessrequires that AI systems be designed
equitably and without biases. Privacy, or “Data Protection and Consent,”
focuses on protecting personal data and obtaining informed consent for its
usage.
These aspects are interconnected, and the intersections of the Venn diagram
illustrate how they overlap. For example, the intersection of transparency and
fairness is called accountability, which emphasizes the importance of trans-
parent and fair AI decision-making. The intersection of transparency and
privacy, called user trust, emphasizes the need for transparency in data use
and protection (Aldoseri, Al-Khalifa, and Hamouda 2023; Bécue, Praça, and
Gama 2021; Malhotra 2018; Mijwil, Aljanabi, and ChatGPT 2023). The inter-
section of fairness and privacy, known as nondiscriminatory data practices,
highlights the need for privacy considerations to align with fairness to avoid
discrimination.
The center intersection of the Venn diagram represents the ideal of respon-
sible AI use that balances transparency, fairness, and privacy. A flowchart
provides a clear, step-by-step guide to embed ethical considerations into AI
development.
The process starts with a commitment to ethical AI development and
defining ethical principles such as transparency, fairness, and privacy. Data
use and AI training guidelines should then be implemented to adhere to these
principles. Regular audits should be conducted to identify and correct biases
and ensure compliance with ethical standards.
Clear lines of responsibility and accountability should be established in AI-
driven decisions. Continuous improvement based on feedback from users and
stakeholders should be a regular practice. The goal is the realization of AI
systems that fully embody ethical principles and ensure the safety and well-
being of all users.
The framework from Figure 1 is discussed in more detail in the next section.
Transparency, Explainability, and Clarity
In developing artificial intelligence, it is essential to prioritize transparency,
explainability, and clarity to ensure ethical development and deployment.
Transparency refers to the accessibility of AI systems and their workings to
users and stakeholders. Explainability, closely linked to transparency, pertains
to the ability of AI systems to be understood and interpreted by human beings,
ideally in non-technical language. Meanwhile, clarity ensures that AI systems’
purposes and outcomes are communicated in a straightforward and under-
standable manner.
The importance of these elements cannot be overstated. They are crucial in
building and maintaining user trust, ensuring that AI systems operate
APPLIED ARTIFICIAL INTELLIGENCE e2463722-5
Figure 1. Transparency, Explainability, and Clarity: A framework for developing AI systems that
embody ethical principles and ensure the safety and well-being of all users.
e2463722-6 P. RADANLIEV
understandably and predictably. Furthermore, transparency and explainability
play a pivotal role in establishing accountability, ensuring that AI developers
and users are held responsible for the outcomes of AI systems (European
Parliament 2023; ISO 2023; McCorduck and Cfe 2004; MeitY 2023; NIST
2023b; Office for Artificial Intelligence 2023).
Fairness and Bias Prevention
Ensuring fairness in AI systems requires creating programs that make deci-
sions without prejudice or partiality (Bender et al. 2021). This necessitates
a conscientious effort to design AI systems that do not perpetuate existing
biases or create new ones (Shu, Zhang, and Yu 2021). However, achieving
fairness in AI poses significant challenges, as these systems often learn from
real-world data, which can be inherently biased.
The intersection of fairness, privacy, and accountability is a complex but
essential consideration. Ensuring fairness often involves careful handling of
sensitive data while also maintaining transparency and accountability in
decision-making processes. This balancing act is critical in mitigating biases
and ensuring that AI systems are equitable and just.
Privacy and Data Protection
Privacy and data protection are critical ethical considerations that must be
considered during AI development. It involves protecting personal and sensi-
tive information from unauthorized access and ensuring that data is used
responsibly. Regulations and standards, such as the General Data Protection
Regulation (GDPR) (GDPR 2018; ICO 2018) in the European Union, play
a significant role in shaping AI ethics by setting strict guidelines for data use.
Privacy, fairness, and user trust are closely linked. Protecting privacy is crucial
in building and maintaining user trust, which is essential for the acceptance
and success of AI systems. Furthermore, ensuring the proper handling of data
is vital for fairness, as data misuse can lead to biased outcomes.
Interconnectedness of Ethical Aspects
The ethical dimensions of AI, including transparency, fairness, and privacy,
are not isolated but deeply interconnected (Partnership on AI 2023; Roberts
et al. 2021). We can visualize this interconnectedness using a Venn diagram
that shows how these aspects overlap and influence each other. For instance,
when transparency and fairness intersect, it leads to accountability. Similarly,
when fairness and privacy overlap, it underscores the need for nondiscrimi-
natory data practices. The idea of responsible AI use is represented by the
central intersection of these aspects in the Venn diagram. This is where all
APPLIED ARTIFICIAL INTELLIGENCE e2463722-7
three principles are balanced, leading to AI systems that are ethical, reliable,
and trustworthy. We can visualize this in Figure 2.
Implementation Strategies
Incorporating ethical considerations into the development of AI requires
a systematic approach. A step-by-step guide for doing this involves first
committing to ethical principles. Next, data use and AI training guidelines
should be implemented to align with these principles. Regular audits are
necessary to detect and correct biases to ensure compliance with ethical
standards.
Establishing clear lines of responsibility and accountability in AI-driven
decisions is also crucial. Continuous improvement, based on feedback from
Figure 2. Interconnected concepts of the Framework for developing AI systems that embody
Transparency, Explainability and Clarity.
e2463722-8 P. RADANLIEV
users and stakeholders, should be an integral part of AI development. The
ultimate goal is to create AI systems that embody ethical principles, ensuring
all users’ safety and well-being.
Responsible AI Frameworks
This section will explore AI frameworks from the European Union, the
United States, and China. Using a detailed flowchart, we will examine how
these frameworks impact each stage of an AI project, from initiation to
post-deployment. Additionally, we will use a Venn diagram analysis to
compare the EU AI Act (Bommasani et al. 2023), US AI Principles
(Tabassi 2023), and China AI Ethics guidelines (Roberts et al. 2021), high-
lighting their unique features and areas of overlap. This approach will
demonstrate how these frameworks share a commitment to ethical stan-
dards, privacy protection, and fairness while also providing distinct per-
spectives on AI development and governance. This analysis is crucial for
understanding the multifaceted nature of global AI frameworks and their
implications for responsible AI practices.
While this paper primarily focuses on theoretical frameworks and policy
analysis, it is important to acknowledge the growing need for empirical
research to substantiate the claims made within the scope of AI ethics. In
response, we introduce two key case studies that provide concrete examples of
how AI ethics frameworks are applied in practice. These case studies were
derived from in-depth analyses of recent AI implementations in both the
Figure 3. Key elements found in global AI frameworks.
APPLIED ARTIFICIAL INTELLIGENCE e2463722-9
healthcare and financial sectors, which serve as critical areas where ethical
considerations are paramount.
The first case study examines the deployment of AI diagnostic systems in
the European healthcare industry, particularly focusing on IBM Watson
Health’s use of AI in oncology diagnostics. By conducting structured inter-
views with healthcare practitioners and reviewing compliance reports, we
observed how the stringent transparency and accountability requirements
mandated by the European Union’s AI Act (European Parliament 2023)
influenced the AI system’s design and operational transparency. This empiri-
cal evidence demonstrates how AI systems were modified to meet regulatory
standards, particularly concerning the explainability of diagnostic recommen-
dations provided to medical professionals and patients.
The second case study involves an empirical assessment of AI fraud
detection systems in the financial services sector within the United States.
Through direct engagement with industry experts and an analysis of
Figure 4. Applied design of responsible and ethical AI practices by integrating global AI frame-
works into different stages of AI development and deployment.
e2463722-10 P. RADANLIEV
internal auditing processes, we explored how JP Morgan’s AI-powered
fraud detection system aligns with the flexible, innovation-driven guidelines
of the NIST AI Risk Management Framework (NIST 2023a; Tabassi 2023).
The study reveals how these frameworks permit adaptive risk management
strategies, providing companies with the autonomy to tailor their ethical
standards while maintaining a balance between innovation and
accountability.
These case studies provide empirical evidence to support the theoretical and
policy analysis discussed in this paper. They illustrate how global AI ethics
frameworks are not just abstract concepts but operational guidelines that have
tangible impacts on AI design, deployment, and compliance. By incorporating
these real-world applications, we aim to bridge the gap between theory and
practice, offering a more robust foundation for understanding the dynamics of
AI governance across various industries.
Figure 5. Different global AI frameworks overlap in some areas but maintain unique characteristics
in others.
APPLIED ARTIFICIAL INTELLIGENCE e2463722-11
Development and Deployment Flowchart
Creating and implementing AI systems is a complex process that requires
compliance with international frameworks to ensure ethical and responsible
outcomes. This section provides a comprehensive flowchart that integrates AI
frameworks from the European Union, the United States, and China, mapping
their application throughout the AI project lifecycle.
The flowchart is based on the EU AI Act (European Parliament 2023), US
AI Principles (NIST 2024c), and China AI Ethics guidelines (Provisions on the
Figure 6. The complexity of AI frameworks across different regions and the potential for collabora-
tion and divergence highlights the need for global collaboration and harmonisation of AI ethics
and regulations.
e2463722-12 P. RADANLIEV
Administration of Deep Synthesis Internet Information Services, Personal
Information Protection Law of the People’s Republic of China PRC 2022). It
begins with initiating an AI project, where objectives are defined and relevant
data is collected. The EU AI Act’s guidelines on ethical data use and transpar-
ency are essential at this stage. The US AI Guidelines come into play as the
project progresses to data processing and AI model development, emphasizing
innovation, fairness, and accountability.
During the validation and testing phase, the model must align with the set
objectives and comply with ethical standards per these frameworks. The
deployment phase sees the integration of China’s AI Ethics guidelines, which
prioritize social harmony, national security, and global cooperation. After
deployment, continuous monitoring and maintenance are essential to ensure
the AI system functions as intended and adheres to ethical standards. This
phase is critical for incorporating feedback and insights, allowing for iterative
improvements based on real-world performance and impact.
Figure 3 illustrates how global AI frameworks are necessary and applicable
at different AI development and deployment stages. This integration ensures
a holistic approach to responsible AI practices that align with global standards
and ethical considerations.
Figure 3 shows the steps involved in developing and deploying AI systems.
It incorporates the latest AI frameworks worldwide, including those from the
EU, the US (NAIAC 2024, NIST 2024a), and China (Interim Measures for the
Management of Generative Artificial Intelligence Services, Personal
Information Protection Law of the People’s Republic of China PRC 2023; Li
2017; Provisions on the Administration of Deep Synthesis Internet
Information Services, Personal Information Protection Law of the People’s
Republic of China PRC 2022; The State Council People Republic of China
2017), and highlights critical points where these frameworks intersect. The
process flow begins with the start of the AI project and moves on to identifying
objectives and collecting relevant data. The EU AI Act’s guidelines may be
considered at this stage. The collected data is then processed, and the AI model
is developed according to the principles outlined in the US AI Guidelines. The
model is then validated and tested to meet the set objectives. Deployment of
the AI system in a real-world environment is the next step, and considerations
from China’s AI Ethics guidelines come into play here. Continuously mon-
itoring and maintaining the AI system post-deployment is essential. The
flowchart also includes a feedback loop that involves revisiting objectives
and processes based on feedback and new insights. Once all the steps are
completed, the AI project cycle ends. The flowchart in Figure 4 ensures
responsible and ethical AI practices by integrating global AI frameworks
into different stages of AI development and deployment.
Figure 4 visualizes how various AI frameworks integrate with the AI devel-
opment and deployment process. Each stage of the AI process is marked, from
APPLIED ARTIFICIAL INTELLIGENCE e2463722-13
the beginning to the end. The EU, the US, and China AI frameworks are
highlighted with individual annotations. The arrows linking these frameworks
to specific stages in the process are designed to avoid text overlap, ensuring
clarity and readability.
The flowchart effectively illustrates the integration of global AI frame-
works into the AI development lifecycle, emphasizing the importance of
considering these guidelines at different stages for responsible AI practices.
The annotations for the AI frameworks from the EU, US, and China are
strategically positioned to avoid obstructing any other flowchart elements.
The connecting arrows are designed with a specific arc to neatly link the
frameworks to their respective stages in the AI process without crossing
over any text.
The Venn diagram represents the AI frameworks from the EU, the US, and
China. Each circle in the diagram represents a different region’s AI framework:
EU AI Act, US AI Principles, and China AI Ethics. The overlaps between the
circles indicate areas of common focus or principles shared between these
frameworks. Individual sections highlight unique aspects of each framework.
This Venn diagram in Figure 5 visually demonstrates how different global
AI frameworks overlap in some areas while maintaining unique characteristics
in others. It reflects the diverse approaches to AI governance and ethics across
these regions.
The Venn diagram in Figure 5 comprehensively analyses the AI frameworks
from the European Union, the United States, and China. It highlights the areas
of collaboration and divergence between the three regions and demonstrates
the complexity of AI frameworks across different regions.
The European Union’s AI Act focuses on human oversight, nondiscrimina-
tion, and regulatory compliance. The US AI Principles emphasize innovation
encouragement, public trust, and open collaboration. China’s AI Ethics prior-
itizes social harmony, national security, and global cooperation. These unique
elements reflect the individual priorities and cultural perspectives of each
region.
The EU and the US share values in transparency and ethical standards. The
EU and China both emphasize privacy protection and ethical standards. The
US and China find common ground in ethical standards and a focus on
fairness. These common areas between the two frameworks indicate the
potential for global collaboration and harmonization of AI ethics and
regulations.
The central overlap in the Venn diagram highlights the areas where the EU,
US, and China share common principles such as ethical standards and privacy
protection. These areas offer opportunities for global collaboration and har-
monization of AI ethics and regulations.
Each region has unique focus areas that reflect its priorities and cultural
perspectives. These divergent approaches could lead to different approaches in
e2463722-14 P. RADANLIEV
AI development and governance. The Venn diagram demonstrates the poten-
tial for collaboration and divergence in the global AI landscape.
The Venn diagram analysis in Figure 6 shows the complexity of AI frame-
works across different regions and the potential for collaboration and diver-
gence. It highlights the need for global collaboration and harmonization of AI
ethics and regulations to ensure that AI development and governance align
with ethical and societal values.
The Venn diagram in Figure 6 shows the EU AI Act i in light blue, the US AI
Principles are in light green, and the China AI Ethics framework in coral.
Real-World Applications of Global AI Frameworks
In order to provide a clearer understanding of how AI frameworks function in
practice, it is essential to examine real-world applications of these frameworks
across different regions. A relevant example can be found in the application of
the European Union’s AI Act within the healthcare sector. The use of AI-
powered diagnostic tools, such as IBM’s Watson Health, was subject to
scrutiny under the EU’s stringent regulations on transparency and explain-
ability. Under the AI Act, companies deploying such AI systems in high-risk
sectors are required to provide detailed documentation of the algorithms used,
as well as explainability mechanisms that allow medical professionals and
patients to understand AI-driven decisions. This regulatory requirement has
led to the modification of AI models to ensure compliance, particularly by
providing more transparent decision-making processes that can be audited by
healthcare regulators.
In contrast, the United States’ more innovation-centric approach, as embo-
died by the NIST AI Risk Management Framework, can be observed in the
deployment of AI in the financial services sector. For instance, JP Morgan’s
AI-powered fraud detection system operates within a framework that empha-
sizes risk mitigation through best practices and industry standards rather than
rigid regulatory oversight. The NIST framework encourages companies to
develop internal policies tailored to their operational risks, allowing for greater
flexibility in AI implementation. As a result, JP Morgan has developed pro-
prietary methods for continuous monitoring and auditing of AI models to
ensure they remain effective while balancing the need for innovation with
ethical considerations.
China’s AI governance, which prioritizes state control and societal
harmony, can be seen in the government’s use of facial recognition
systems for public security. The deployment of such systems, governed
by China’s Provisions on the Administration of Deep Synthesis Internet
Information Services, Personal Information Protection Law of the
People’s Republic of China (PRC) (2022), illustrates how the state
leverages AI under a framework that prioritizes national security. In this
APPLIED ARTIFICIAL INTELLIGENCE e2463722-15
case, AI technologies are used to monitor public spaces, but their ethical
implications, particularly in terms of privacy, are handled within
a governance model that differs significantly from those in Western
democracies. The Chinese government’s emphasis on AI as a tool for
societal stability underscores the unique application of their framework in
practice.
These examples highlight the varied approaches of AI frameworks across
different regions and sectors, demonstrating how global AI policies are shaped
by contextual factors and applied in practical, high-impact scenarios. By
examining such real-world cases, we can better understand the strengths and
limitations of these frameworks and the challenges in harmonizing AI ethics
on a global scale.
Specific Recommendations for Real-World Applications of Global AI Frameworks
Include
Technical Solutions for Embedding Ethical Principles in AI Systems. Embedding
ethical principles such as fairness, transparency, and accountability into AI
systems requires sophisticated algorithmic approaches that ensure these objec-
tives are met without compromising the system’s performance. Algorithmic
fairness can be addressed using methods such as differential fairness and fair
representation learning. For instance, algorithms like the Fair Representation
Learning (FRL) model aim to mitigate bias by transforming raw data into
a latent representation that is invariant to sensitive attributes, such as race or
gender, without losing important predictive power. The FRL method applies
adversarial learning to ensure the model cannot easily infer sensitive attri-
butes, thus reducing bias while maintaining accuracy. This can be particularly
useful in sectors like finance, where historical biases in credit scoring datasets
often lead to unfair outcomes. Incorporating these fairness constraints during
the model training phase ensures that discriminatory patterns in the data are
not propagated by the AI system.
Transparency is enhanced through the use of explainable AI (XAI) tech-
niques. One common approach is the implementation of Local Interpretable
Model-agnostic Explanations (LIME), which provides users with interpreta-
ble approximations of complex models, enabling end-users and auditors to
understand and evaluate individual predictions. LIME works by perturbing
input data and observing how changes impact predictions, thus constructing
simpler, interpretable models locally around specific instances. This method is
particularly valuable in high-stakes fields like healthcare, where understanding
the rationale behind AI-driven diagnoses is critical for building trust and
accountability. For example, LIME has been effectively applied in medical
imaging to explain how AI systems identify tumor regions, offering transpar-
ency to both clinicians and patients.
e2463722-16 P. RADANLIEV
Obtaining Real-World Probabilistic Data for Legislation. To create more robust
AI legislation that addresses real-world challenges, probabilistic data collec-
tion is necessary. A critical solution lies in data-driven simulations that use
real-world probabilistic distributions of AI outcomes across various domains.
These simulations can leverage Bayesian inference models to analyze the
probability of ethical failures, such as biased decisions or transparency
breaches, under different regulatory scenarios. For example, Bayesian models
can assess the likelihood of biased outputs in loan approval systems based on
varying regulatory constraints, allowing policymakers to quantitatively evalu-
ate the trade-offs between stringent regulation and innovation. By incorporat-
ing such probabilistic assessments, policymakers can develop legislation
grounded in empirical evidence, ensuring that ethical guidelines are both
practical and enforceable in diverse sectors.
Moreover, quantitative data collection from real-world AI deployments
could utilize techniques such as differential privacy to protect sensitive
information while still gathering meaningful insights. For instance, in health-
care, collecting large-scale patient data from AI diagnostic tools while main-
taining patient privacy can be achieved through differential privacy algorithms
that introduce noise into datasets, ensuring that individual records cannot be
re-identified. This allows regulators to gather accurate statistics on AI system
performance, such as prediction accuracy and error rates, without violating
privacy laws. These real-world data points can then be used to fine-tune
legislative frameworks to ensure they are reflective of practical AI use and
compliant with privacy standards.
Algorithmic Solutions to Ensure Fair and Ethical AI for End-Users. To ensure
AI systems are perceived as fair and ethical by end-users, several algorithmic
approaches can be integrated into the development lifecycle. One promising
method is the use of fairness constraints in model optimization, such as
Equalised Odds and Demographic Parity. The Equalised Odds algorithm
ensures that an AI system has equal true positive and false positive rates across
different demographic groups, ensuring that no group disproportionately
benefits or suffers from the system’s decisions. This technique has been
successfully implemented in judicial systems where AI models are used for
bail and sentencing recommendations, reducing the racial disparities com-
monly observed in earlier models.
Fairness-aware learning algorithms can also be embedded into machine
learning pipelines to monitor and adjust for bias during the training process.
For example, the Fairness through Awareness (FTA) framework adjusts
decision boundaries within models to ensure that similar individuals are
treated similarly, thereby reducing unfair bias. This algorithm calculates dis-
tances in a fairness-sensitive space and ensures that individuals who are close
in this space receive similar predictions. This has been applied in hiring
APPLIED ARTIFICIAL INTELLIGENCE e2463722-17
algorithms to ensure that applicants with similar qualifications, regardless of
demographic attributes, are treated equitably.
Furthermore, end-user engagement with AI systems can be improved
through interactive transparency mechanisms. For instance, counterfactual
explanations can be used to provide users with actionable insights into how
decisions could change if certain inputs were modified. In credit scoring
systems, for example, a counterfactual explanation might inform a user that
their loan was denied due to a low credit score and suggest specific steps, such
as reducing credit card debt, that would lead to approval. By providing users
with clear, actionable insights, these systems not only increase trust but also
empower users to engage more meaningfully with AI-driven decisions.
Algorithmic Accountability and Continuous Monitoring. To maintain ongoing
fairness and ethical standards, continuous monitoring of AI systems is essen-
tial. This can be achieved through algorithmic auditing frameworks that
regularly assess AI systems for adherence to ethical principles post-
deployment. Post-hoc fairness auditing tools, such as AI Fairness 360
(AIF360), provide an open-source toolkit that measures and mitigates bias
in deployed models. These tools can be integrated into AI governance pro-
cesses, ensuring that models remain fair and unbiased as they encounter new
data in real-world environments. AIF360 evaluates fairness through multiple
metrics, such as disparate impact and statistical parity, and enables continuous
recalibration of models to maintain ethical performance.
Incorporating algorithmic accountability systems with real-time feedback
loops ensures that biases introduced by shifts in data distributions (data drift)
are swiftly detected and mitigated. Techniques such as drift detection algo-
rithms, including ADWIN (Adaptive Windowing), continuously monitor the
performance of AI models and trigger retraining when significant deviations
from expected behavior are detected. By automating the detection of ethical
breaches and recalibrating models in response, these systems ensure that AI
remains both effective and ethically compliant over time.
Comparative Venn Diagram Analysis
This section uses a Venn diagram analysis to compare the AI frameworks of
the European Union (EU), the United States (US), and China. The diagram
represents each framework, highlighting their unique features and areas of
overlap, revealing both collaborative potentials and divergent approaches.
The EU’s AI Act prioritizes human oversight, nondiscrimination, and strict
regulatory compliance. It reflects the EU’s emphasis on protecting citizens’
rights in the digital age. The US AI Principles prioritize fostering innovation,
ensuring public trust, and promoting open collaboration. This mirrors the
US’s emphasis on market-driven and innovation-led AI development. On the
e2463722-18 P. RADANLIEV
other hand, China’s AI Ethics framework emphasizes the importance of social
harmony, national security, and global cooperation. It reflects China’s
approach to balancing technological advancement with social stability and
state security.
The intersections of these frameworks in the Venn diagram highlight
shared principles and potential areas for international cooperation. For exam-
ple, the EU and the US emphasize ethical standards and transparency. The EU
and China share a common focus on privacy protection and the ethical use of
AI. The US and China converge on encouraging ethical standards and fairness
in AI.
At the central intersection of the Venn diagram, where all three frame-
works overlap, lies a shared commitment to ethical standards, privacy
protection, and ensuring fairness. This common ground suggests opportu-
nities for global collaboration and the harmonization of AI ethics and
regulations.
However, each framework’s divergent aspects reflect each region’s
varying priorities and cultural perspectives. These differences could
lead to distinct approaches in AI development and governance globally.
Therefore, the Venn diagram highlights the potential for collaboration
and underscores the need to understand and respect diverse perspectives
in the global AI landscape.
Policy Frameworks
This section expands into a thorough analysis of AI Ethics Policy
Frameworks worldwide. It covers significant regions such as the European
Union, the United States, China, Canada, Japan, India, and Australia.
Figure 7 presents the data in Cartesian graphs, comparing these frame-
works across four key ethical dimensions: Transparency, Accountability,
Fairness, and Privacy. Each framework is evaluated in detail and rated,
providing an insightful understanding of how countries prioritize these
dimensions in their AI policies. This graph highlights each framework’s
unique priorities and focus areas and emphasizes the diversity and com-
monalities in global approaches to AI ethics. Policymakers can leverage
these insights to identify areas for improvement and develop comprehen-
sive, ethically aligned AI policies. Moreover, the section expands upon this
analysis to include additional critical dimensions such as human oversight
and national security, broadening the scope to encompass broader socio-
political implications of AI technology. Figure 8 analyses global trends and
presents a roadmap for harmonization in AI ethics, advocating for ongoing
international dialogue and cooperation to foster a responsible and ethical
global AI ecosystem.
APPLIED ARTIFICIAL INTELLIGENCE e2463722-19
Criteria for Rating AI Ethics Policy Frameworks Across Countries
The ratings for each country’s AI ethics policy framework were based on
a detailed analysis of public documents, policy white papers, regulatory guide-
lines, and academic literature. The scores for each dimension, transparency,
accountability, fairness, privacy, human oversight, ethical standards,
Figure 7. Cartesian graph - Comparison of Al Ethics Policy Frameworks Across Countries.
Figure 8. Cartesian graph of Global AI Ethics Policy Frameworks: a tool for policymakers to
understand the priorities of different countries regarding AI ethics.
e2463722-20 P. RADANLIEV
innovation encouragement, and national security, were evaluated according to
the following specific criteria:
Transparency (1–10 Scale)
Transparency was assessed by the extent to which each country’s framework
mandates openness regarding the design, implementation, and decision-
making processes of AI systems.
●9-10: Countries with explicit, legally enforced transparency requirements
for AI systems, including mandates for explainability and public account-
ability (e.g., the European Union).
●6-8: Countries that encourage transparency but do not mandate it as
a legal requirement across all AI applications (e.g., the United States).
●4-5: Countries where transparency is mentioned in policies but with few
practical enforcement mechanisms (e.g., China).
●1-3: Minimal or no formal focus on transparency in the AI framework.
Accountability (1–10 Scale)
Accountability measures the robustness of legal and regulatory mechanisms
that hold developers, companies, and governments responsible for the out-
comes of AI systems.
●9-10: Countries with well-defined liability frameworks that assign clear
responsibility to AI developers or operators (e.g., the EU AI Act).
●6-8: Countries where accountability is encouraged through voluntary
compliance frameworks but lacks mandatory enforcement (e.g., the US
with the NIST AI Risk Management Framework).
●4-5: Countries with vague accountability measures, often handled at the
discretion of private entities or lacking centralized regulation (e.g., Japan).
●1-3: Countries where accountability frameworks are non-existent or still
in early development phases.
Fairness (1–10 Scale)
Fairness was evaluated based on how well a country’s policy framework
addresses bias in AI algorithms and ensures equitable outcomes across demo-
graphic groups.
●9-10: Countries with explicit fairness requirements for AI systems, man-
dating fairness audits and bias mitigation techniques (e.g., Canada, EU).
●6-8: Countries that encourage fairness but with less stringent or optional
auditing practices (e.g., the US).
●4-5: Countries where fairness is an aspirational goal with limited practical
implementation (e.g., India).
APPLIED ARTIFICIAL INTELLIGENCE e2463722-21
●1-3: Minimal or no focus on fairness in AI regulation.
Privacy (1–10 Scale)
Privacy was assessed by how each country’s framework protects user
data in the context of AI and how it aligns with global standards like
the GDPR.
●9-10: Countries with robust privacy regulations, including explicit rules
on AI data use (e.g., EU, Canada).
●6-8: Countries with general data privacy laws but limited AI-specific
guidelines (e.g., the US, Japan).
●4-5: Countries with privacy regulations that are inconsistently applied or
underdeveloped in relation to AI (e.g., India, Australia).
●1-3: Countries with minimal focus on privacy in AI contexts, or where
data protection laws are not enforced effectively (e.g., China).
Human Oversight (1–10 Scale)
This criterion assessed the role of human oversight in AI decision-
making, particularly in high-risk sectors like healthcare or autonomous
vehicles.
●9-10: Countries mandating human oversight in high-risk AI decisions,
ensuring human intervention in critical areas (e.g., EU AI Act).
●6-8: Countries that recommend but do not legally enforce human over-
sight (e.g., the US).
●4-5: Countries where human oversight is mentioned, but enforcement
mechanisms are vague or absent (e.g., Japan, India).
●1-3: Little to no emphasis on human oversight in AI policy frameworks.
Ethical Standards (1–10 Scale)
Ethical standards were scored based on how well a country’s AI policies adhere
to global ethical frameworks (such as UNESCO’s AI ethics recommendations)
and promote ethical AI development.
●9-10: Countries with a clearly defined, internationally aligned ethical
framework for AI (e.g., the EU, Canada).
●6-8: Countries with ethical guidelines for AI but limited in scope or
enforcement (e.g., Japan, the US).
●4-5: Countries that mention ethical AI but lack a coherent, enforceable
framework (e.g., India).
●1-3: Minimal or no formal focus on AI ethics in public policy.
e2463722-22 P. RADANLIEV
Innovation Encouragement (1–10 Scale)
This criterion measured the balance between promoting AI innovation and
enforcing ethical guidelines. Countries that foster innovation while maintain-
ing a robust ethical framework scored higher.
●9-10: Countries with innovation-centric policies that support AI research
and development while integrating ethical guidelines (e.g., US, Canada).
●6-8: Countries with strong innovation policies but less rigorous ethical
enforcement (e.g., Japan).
●4-5: Countries where innovation is promoted but at the cost of ethical
standards (e.g., China).
●1-3: Countries where innovation in AI is stifled due to excessive regula-
tion or a lack of resources (e.g., minimal focus).
National Security (1–10 Scale)
National security was evaluated based on how countries incorporate AI within
their national security strategies, including defense, cybersecurity, and
surveillance.
●9-10: Countries where AI plays a significant role in national security
frameworks, with clear policies on military AI, surveillance, and cyber
defense (e.g., China, the US).
●6-8: Countries that include AI in national security policies but with fewer
explicit regulations on its use in defense (e.g., Australia, Japan).
●4-5: Countries with some mention of AI in national security contexts but
lacking concrete policies (e.g., India).
●1-3: Minimal focus on AI for national security, or policies that are still in
early development (e.g., the EU).
Justication for the Selection of Criteria and Scores
The chosen criteria for evaluating AI ethics policy frameworks, transparency,
accountability, fairness, privacy, human oversight, ethical standards, innova-
tion encouragement, and national security, were carefully selected to reflect
the core dimensions that are essential for ethically sound, socially beneficial,
and technologically responsible AI systems. These dimensions are well-
established in policy discourse and academic literature as the pillars of AI
ethics and governance, ensuring that AI development aligns with societal
values and mitigates potential harms.
Transparency
Transparency is a cornerstone of AI ethics, as highlighted in both academic
and regulatory discussions (Floridi et al. 2018). Transparent AI systems allow
APPLIED ARTIFICIAL INTELLIGENCE e2463722-23
stakeholders to understand how decisions are made and ensure accountability.
The choice of transparency as a criterion is supported by regulatory frame-
works such as the European Union’s GDPR and AI Act, which place explicit
demands on AI systems to be explainable and open to public scrutiny. Studies
have shown that lack of transparency is one of the main causes of public
distrust in AI systems (Wachter, Mittelstadt, and Russell 2023). Therefore,
countries with clear legal mandates for transparency received higher scores,
while those with voluntary or vague transparency guidelines scored lower.
Accountability
Accountability ensures that AI developers, operators, and users are held
responsible for the outcomes produced by AI systems. This criterion is
justified by the recognition in the literature that without clear accountability
structures, it becomes difficult to address failures or harms caused by AI
systems (Mittelstadt 2019). The EU AI Act introduces comprehensive provi-
sions that assign legal responsibility, providing a strong model for account-
ability. Countries like the United States, with voluntary compliance through
frameworks like the NIST AI Risk Management Framework, received inter-
mediate scores due to the lack of enforceability. The necessity of accountability
is also a major theme in academic literature, particularly in the context of
complex AI systems where multiple stakeholders are involved in the design
and deployment (de Bruin and Floridi 2017; Floridi et al. 2018; Turilli and
Floridi 2009).
Fairness
Fairness in AI systems addresses concerns about bias and discrimination,
which are well-documented issues in AI applications (Binns 2018).
Countries with explicit fairness requirements in their AI policies, such as the
European Union and Canada, received higher scores because their frameworks
mandate fairness audits and bias mitigation practices. Literature on fairness in
AI often points to the limitations of algorithmic systems to ensure equitable
outcomes across demographic groups without specific regulatory intervention
(Du 2023; IBM 2018). Nations with minimal or non-enforceable fairness
provisions, such as India and China, scored lower due to the absence of robust
bias-mitigation mechanisms.
Privacy
Privacy is a critical concern in AI, especially in systems that rely on vast
amounts of personal data. The GDPR in the EU sets a high global benchmark
for data protection and privacy, justifying the high score for the EU in this
dimension. In contrast, countries like the United States, where privacy regula-
tions such as HIPAA are domain-specific and not universally applicable to AI
systems, scored lower (HIPAA 1996). Privacy as a criterion is grounded in the
e2463722-24 P. RADANLIEV
principle that ethical AI must protect individuals’ rights to control their data,
a concern that is pervasive in academic and policy literature (Mittelstadt 2019).
Human Oversight
Human oversight in AI decision-making is essential to prevent over-reliance
on automated systems, particularly in critical sectors like healthcare and law
enforcement (European Parliament 2023). The EU AI Act again leads the way
by mandating human oversight in high-risk AI applications. The literature
emphasizes the importance of preserving human judgment in AI-assisted
decision-making, especially in cases involving moral or legal consequences
(Jobin, Ienca, and Vayena 2019). Countries that recommend but do not
mandate human oversight scored lower, as voluntary oversight often fails in
real-world applications, particularly where operational efficiency is prioritized
over human intervention.
Ethical Standards
Ethical standards are increasingly seen as vital for aligning AI development
with human values. UNESCO’s Recommendation on the Ethics of AI and
similar initiatives by the OECD have provided blueprints for ethical AI,
focusing on principles such as beneficence, non-maleficence, autonomy, and
justice (UNESCO 2023). Countries like Canada and the EU, which have
adopted comprehensive ethical guidelines, scored highly. These standards
are crucial for ensuring that AI operates within moral and legal boundaries.
In contrast, countries that lack specific ethical frameworks for AI, such as
India and China, received lower scores, reflecting the underdevelopment of
ethical considerations in their AI policies.
Innovation Encouragement
The balance between encouraging AI innovation and enforcing ethical stan-
dards is a key concern for policymakers (Brynjolfsson and Mcafee 2014; Evans
2015). Countries like the United States scored high in this dimension due to
their innovation-centric policies, such as the NIST AI Risk Management
Framework, which promotes industry-led solutions and fosters a favorable
environment for AI research and development. The literature supports the
notion that innovation thrives when there is flexibility and minimal regulatory
overhead, but with the caveat that ethical guardrails must not be neglected
(Bostrom and Yudkowsky 2014). Countries that focus excessively on regula-
tion, potentially stifling innovation, or that lack sufficient incentives for AI
research, scored lower.
National Security
National security considerations, particularly regarding the development of
autonomous weapons systems (AWS) and AI-enhanced cybersecurity, are
APPLIED ARTIFICIAL INTELLIGENCE e2463722-25
becoming a critical component of AI policy (Singer 2009). Countries like the
US and China, where AI plays a substantial role in national defense strategies,
scored highly. The literature on autonomous systems highlights the impor-
tance of regulating AI to prevent unintended consequences in military appli-
cations (Singer 2009). Countries that have not yet integrated AI into their
national security strategies or have underdeveloped AI governance in this area,
such as the EU, scored lower.
Justification for Scoring. The scores for each dimension were derived from
a combination of the following sources:
●Policy Documents and Regulations: Key regulatory frameworks such as
the EU AI Act (Bommasani et al. 2023; European Parliament 2023, FACT
SHEET: Biden-Harris Administration Announces New Actions to
Promote Responsible AI Innovation That Protects Americans’ Rights
and Safety | The White House, 2023; Mozumder et al. 2022), GDPR
(GDPR 2018; ICO 2018), NIST AI Risk Management Framework
(NIST 2024a, 2024c; Tabassi 2023), and national AI strategies were
directly analyzed to assess the strength and comprehensiveness of each
country’s AI governance mechanisms.
●Academic Literature: Foundational texts on AI ethics, fairness, transpar-
ency, and accountability were referenced to establish baseline expecta-
tions for what constitutes best practices in each dimension (Binns 2018;
Floridi et al. 2018; Jobin, Ienca, and Vayena 2019).
●Real-World Case Studies: Examples of AI implementation in various
sectors, including healthcare, finance, and national security, were exam-
ined to contextualize the practical impacts of each policy framework.
The criteria were selected to ensure a comprehensive assessment of each
country’s approach to AI ethics, focusing on both regulatory stringency and
the practical application of ethical principles. Each score reflects the extent to
which the country’s framework addresses key challenges associated with AI
governance, ensuring a balanced evaluation that is informed by both policy
analysis and academic insights.
Comparative Analysis Using Cartesian Graphs
This section provides a detailed comparative analysis of AI Ethics Policy
Frameworks from a global perspective. The analysis covers major regions
such as the European Union, the United States, China, Canada, Japan, India,
and Australia. The study examines these frameworks against key ethical
dimensions, including Transparency, Accountability, Fairness, and Privacy,
using a series of Cartesian graphs.
e2463722-26 P. RADANLIEV
Each framework is rated on a scale of 1 to 10 across these dimensions. The
Cartesian graph format helps to visualize how each country’s framework
measures up in these critical areas. For instance, the EU’s framework prior-
itizes privacy and accountability, reflected in its high scores in these areas. On
the other hand, the US framework might score higher on transparency because
of its focus on open data and innovation. China’s framework, emphasizing
social harmony and national security, might have different strengths and
weaknesses.
This comparative analysis provides insights into the priorities and focus
areas of different countries and highlights the diversity and commonalities in
approaches to AI ethics globally. The graphical representation aids in under-
standing the complexities of each framework, offering a clear view of how
nations are navigating the ethical landscape of AI development.
The Cartesian graph in Figure 7 compares AI Ethics Policy Frameworks
across countries, such as the EU, the US, China, Canada, Japan, India, and
Australia. It evaluates them on Transparency, Accountability, Fairness, and
Privacy.
Figure 7 presents an overview of how countries prioritize various elements
of AI ethics in their policy frameworks. The four aspects used for rating are
transparency, accountability, fairness, and privacy. Each aspect is rated on
a scale from 1 to 10.
The Blue Line represents transparency, which reflects how policies are
openly communicated and implemented. The Green Line indicates account-
ability, measuring the extent to which AI developers and users are held
accountable for their systems as per the frameworks. The Red Line represents
fairness, measuring the extent to which the policies ensure fair and unbiased
AI systems. Lastly, the Purple Line shows the importance of user privacy and
data protection in the policies.
The graph offers valuable insights into the global landscape of AI ethics. It
highlights similarities and differences in national approaches to regulating AI
that can help inform future policy development. The ratings of each aspect for
each country can help policymakers identify areas for improvement in their
policies.
This graph provides a tool for policymakers to understand the priorities of
different countries regarding AI ethics. By focusing on transparency, account-
ability, fairness, and privacy, policymakers can create policies that address the
concerns of stakeholders and help establish ethical guidelines for AI develop-
ment and use.
Extended Analysis and Global Implications
In this section, we expand into an evaluation matrix that incorporates addi-
tional dimensions that are increasingly relevant to AI ethics. These dimensions
APPLIED ARTIFICIAL INTELLIGENCE e2463722-27
include the roles of human oversight and national security, reflecting AI
technology’s broader socio-political implications.
Including human oversight highlights the necessity for human intervention
and judgment in AI systems, a principle strongly supported by the EU frame-
work. Conversely, national security is crucial in the frameworks of countries
like China and the US, where AI’s involvement in defense and intelligence is
a pivotal consideration.
This comprehensive analysis emphasizes the global trends in AI policy
frameworks and the potential for harmonizing AI ethics. The section
explores the possibility of converging principles and standards despite
each region’s diverse cultural, political, and social contexts. While complete
uniformity may be unattainable, the potential for international collabora-
tion and consensus-building on core principles is significant. Such harmo-
nization could facilitate the establishment of universally accepted norms
and standards, ensuring that AI development aligns with ethical and socie-
tal values.
The analysis of similarities and differences in Figure 8 concludes that we
need continued dialogue and cooperation among nations to cultivate
a responsible and ethical global AI ecosystem.
The Cartesian graph in Figure 8 includes lines representing concepts
inspired by earlier Venn diagrams that explored the issues surrounding AI
governance.
The lines on the graph represent different aspects of AI ethics that
have been identified as necessary. The blue line represents transparency,
which refers to the openness and clarity in AI policy communication.
The green line represents accountability, which signifies the extent of
responsibility in AI development and use. The red line represents fair-
ness, which denotes the importance of ensuring unbiased AI systems.
The purple line represents privacy, which highlights the importance of
user privacy and data protection. The orange line represents human
oversight, which signifies humans’ involvement and oversight in AI
processes. The brown line represents ethical standards, which means
adherence to ethical guidelines in AI. The pink line represents innova-
tion encouragement, which reflects support for innovative AI develop-
ment. The gray line represents national security, emphasizing AI’s role
in national security.
These lines provide a comprehensive view of how different countries
address multiple facets of AI ethics in their policies. They reflect a broader
range of considerations in the global discourse on AI governance. By con-
sidering the different aspects of AI ethics, policymakers can create fair, trans-
parent, and accountable policies while encouraging innovation and protecting
user privacy and data. This is essential to building trust in AI and ensuring that
it is used to benefit society.
e2463722-28 P. RADANLIEV
Strategies to Mitigate Bias
This section expands into the issue of bias in Artificial Intelligence (AI)
systems and its significant impact on fairness and effectiveness. We introduce
Figure 8, a network diagram visually representing various interconnected
strategies crucial for mitigating bias. These include ensuring data diversity to
avoid biased AI models, conducting regular audits to detect and correct biases,
employing bias detection tools, and emphasizing the importance of algorith-
mic transparency. In addition, we emphasize the importance of integrating
diverse development teams and providing ethical AI training to reduce uncon-
scious bias in AI design and development. The section then transitions to
Figure 10, which provides a more detailed network representation, highlight-
ing the synergies and dependencies among these strategies. This section
demonstrates how a collective, multifaceted approach, comprising both tech-
nical and organizational measures, is vital for developing AI systems that are
equitable, fair, and aligned with ethical standards. It is essential to note that
mitigating bias in AI is an ongoing process that requires continuous vigilance
and adaptation to evolving AI technologies and societal norms.
Comprehensive Mitigation Strategies
Bias in AI systems is of utmost importance, as it can significantly impact the
fairness and effectiveness of these technologies. This section reviews various
strategies aimed at mitigating bias, presented as a network diagram showcas-
ing the interconnectedness and collective importance of these strategies.
The network diagram encompasses a range of approaches, each linked to
demonstrate how they complement and reinforce one another. Key strategies
include ensuring data diversity and using datasets representing all relevant
demographics to prevent biased AI models. Regular audits are crucial to
identifying and addressing biases that may develop over time. Additionally,
the network emphasizes the importance of using bias detection tools, which
employ specialized algorithms to uncover and address biases in AI systems.
The diagram in Figure 9 emphasizes the significance of collectively imple-
menting these strategies, indicating that the most effective approach to miti-
gating bias involves a multifaceted effort. This includes technical solutions and
organizational and procedural measures to ensure that AI systems are devel-
oped and operated in a manner that minimizes bias and promotes fairness.
The diagram in Figure 9, shows the connections and assigned weights
between different strategies to mitigate bias in AI. The thickness of the lines
corresponds to the strength of the connection, and the weights are labeled on
the diagram. The strategies are arranged circularly to emphasize their inter-
connectedness and importance. By implementing them collectively, the risk of
APPLIED ARTIFICIAL INTELLIGENCE e2463722-29
Figure 9. The significance of a collective implementation of AI Strategies: connections and Weights
in Strategies to Mitigate Bias in Al (Colour-coded by Strength).
Figure 10. Strategies to Mitigate AI Bias.
e2463722-30 P. RADANLIEV
bias in AI systems can be significantly reduced, leading to more equitable and
trustworthy AI solutions.
One key strategy is to ensure data diversity and representation. This
involves using diverse data representing all relevant demographics to avoid
biased models in AI systems. Regular audits are also essential to identify and
rectify any biases that may have crept in over time. Bias detection tools are
another essential strategy. These tools utilize specialized software to detect
biases in AI algorithms, which can then be corrected. Diverse development
teams can also help minimize unconscious biases in designing and developing
AI systems.
Providing ethical AI training to AI professionals is another way to mitigate
bias in AI. This training educates AI professionals on ethical considerations
and avoiding bias. Algorithmic transparency is also crucial to reducing bias in
AI. By making the workings of AI algorithms transparent, biases can be
identified and corrected more quickly. Involving various stakeholders, includ-
ing those from underrepresented groups, to provide feedback on AI systems
and their outputs can also significantly reduce bias. Finally, continuously
monitoring AI systems is essential to quickly identify and address any biases
that may emerge over time. Collectively implementing these strategies can
significantly reduce the risk of bias in AI systems, leading to more equitable
and trustworthy AI solutions.
The flowchart in Figure 10 outlines a network representation of strategies
for mitigating bias in AI. The enhanced diagram provides more context and
displays the connections between these strategies, providing a comprehensive
approach to addressing this issue.
As shown in Figure 9, data diversity is one of the primary strategies to
mitigate bias in AI. It is essential to ensure that data is collected from diverse
demographics. This strategy is linked to bias detection tools, emphasizing the
importance of having a wide range of data to identify and correct biases.
Regular audits are another crucial component of mitigating bias in AI systems.
Periodic reviews of AI systems for biases are required and are connected to
continuous monitoring. This highlights the need for ongoing assessments to
ensure the AI system remains unbiased.
Using bias detection tools is also essential in mitigating bias in AI systems.
This strategy links to algorithm transparency, underscoring detection tools’
role in making AI decisions more straightforward. Ensuring that the AI
algorithm is transparent makes it easier to identify and correct any potential
biases.
Diverse teams are also vital to mitigating bias in AI systems. Having diverse
backgrounds in development teams can significantly reduce unconscious
biases. This strategy is connected to ethical training, showing the importance
of diverse perspectives in ethical AI development. This ensures that the AI
system is developed ethically and unbiasedly.
APPLIED ARTIFICIAL INTELLIGENCE e2463722-31
Ethical training is another crucial strategy in mitigating bias in AI systems.
Training on ethical AI development practices is vital, and this connects to
stakeholder feedback. This illustrates the role of ethical considerations in
incorporating diverse viewpoints, ensuring that the AI system is developed
with the interests of all stakeholders in mind.
Making AI algorithms’ decisions clear is another essential strategy for
identifying biases. This is connected to regular audits, highlighting the need
for transparency in ongoing assessments. Ensuring that the AI algorithm’s
decisions are transparent makes it easier to identify and correct any potential
biases.
Incorporating feedback from all groups, including underrepresented ones,
is essential in developing an unbiased AI system. This relates back to data
diversity, emphasizing the role of inclusive feedback in ensuring diverse data
representation. Diverse feedback and ongoing surveillance for emerging biases
is necessary to mitigate bias in AI systems. This relates to diverse teams,
underscoring the need for continuous oversight by teams with varied back-
grounds and perspectives. Diverse teams make it easier to identify and correct
potential biases.
The network diagram in Figure 9 illustrates the interconnected nature of
these strategies, showing how each contributes to a comprehensive approach
to mitigating bias in AI systems. By implementing these strategies, AI systems
can be developed more ethically and unbiasedly, with the interests of all
stakeholders in mind.
Ethical Training and Diverse Teams
Providing ethical training to AI professionals is crucial to making them
aware of potential biases and fostering an ethical culture in AI devel-
opment. This training should cover the ethical implications of AI, the
significance of diversity in datasets, and ways to detect and mitigate
bias.
Forming diverse teams is also a vital strategy. Teams composed of people
from diverse backgrounds bring unique perspectives to the AI development
process, which can help identify biases that a more homogeneous group may
overlook. The diversity here refers to demographic factors and variations in
expertise, experience, and viewpoints.
By combining ethical training, diversity in teams, and technical strate-
gies such as data diversity and regular audits, AI systems can be devel-
oped in a way that is more equitable, fair, and aligned with ethical
standards. Countering bias is a continuous process that necessitates
ongoing attention and adaptation as AI technologies progress.
e2463722-32 P. RADANLIEV
Emerging AI Technologies and Their Ethical Challenges
Large language models (LLMs), such as GPT-3 and GPT-4, exemplify the
growing power of generative AI. These models are trained on vast datasets,
often scraping data from the internet, which raises significant concerns over
data provenance, copyright infringement, and privacy violations. The opa-
que nature of these models complicates efforts to ensure that they are free from
biases present in the training data, such as discriminatory language, misinfor-
mation, or unintentional perpetuation of harmful stereotypes. Despite
employing fine-tuning and debiasing techniques, these models are still prone
to producing biased outputs due to the inherent limitations of the training
data and the probabilistic nature of their generation processes. For instance,
techniques such as reinforcement learning from human feedback (RLHF)
have been deployed to mitigate harmful outputs, but they remain insufficient
in addressing the deeper systemic biases embedded within the underlying
datasets. This calls for more sophisticated techniques, such as adversarial
training, where adversarial examples are used to iteratively refine models
and expose hidden biases. Additionally, federated learning presents
a promising approach for enhancing the ethical training of LLMs by allowing
models to learn from decentralized, anonymized data, thus reducing the
ethical risks associated with data centralization and privacy violations.
Another critical issue with LLMs lies in their ability to produce convincing
but factually incorrect or hallucinatory outputs. This problem, often referred
to as the “hallucination problem,” presents ethical challenges in high-stakes
domains such as healthcare or law, where accurate information is paramount.
Current mitigation strategies include truth-verification models that cross-
check generated content against verified databases and automated fact-
checking systemsintegrated into the model’s inference pipeline. However,
these methods are still evolving and are far from fully resolving the issue.
Ethical frameworks for LLM deployment must, therefore, include rigorous
post-deployment monitoring and real-time validation mechanisms to ensure
the integrity of the outputs, particularly in applications where misinformation
could have profound societal impacts.
Autonomous systems, including autonomous vehicles, drones, and
robotic systems, pose additional ethical challenges related to safety,
accountability, and decision-making autonomy. A key ethical dilemma
arises in the context of autonomous decision-making in unpredictable
environments. For instance, in the case of autonomous vehicles, ethical
frameworks must account for the so-called trolley problem scenarios,
where the system must make life-and-death decisions in the event of an
unavoidable accident. Traditional rule-based ethical systems, such as
deontological or utilitarian approaches, often fail to provide clear solu-
tions in these nuanced scenarios. Consequently, emerging solutions
APPLIED ARTIFICIAL INTELLIGENCE e2463722-33
involve the use of ethical AI algorithms like multi-objective optimiza-
tion, which allows systems to balance competing ethical principles – such
as minimizing harm and respecting human autonomy – by assigning
dynamic weights to different ethical outcomes based on real-time envir-
onmental factors.
Furthermore, accountability in autonomous systems presents a unique
challenge, especially in cases where systems operate with minimal human
oversight. Explainable AI (XAI) plays a critical role here, enabling transpar-
ency in decision-making processes by providing interpretable insights into
how the system reached a specific decision. Techniques such as attention
mechanisms and saliency maps can be employed to highlight the features that
most influenced an autonomous system’s decision, making it easier for reg-
ulators and auditors to understand and assess the fairness and safety of these
decisions. However, the effectiveness of XAI in highly complex, real-time
autonomous systems remains limited, necessitating the development of causal
inference models that can provide a more comprehensive understanding of
decision-making pathways and their underlying ethical implications.
The issue of algorithmic accountability in autonomous weapons systems
(AWS) presents perhaps the most acute ethical challenge. The development
and deployment of AWS raise profound concerns over autonomous lethality
—the ability of a system to make life-or-death decisions without human
intervention. Current discussions on international AI governance focus on
the need to restrict the deployment of AWS through legally binding treaties,
but enforcement mechanisms remain elusive. From a technical standpoint,
one proposed solution involves embedding human-in-the-loop (HITL)
mechanisms that ensure critical decisions, particularly those involving the
use of lethal force, require human validation before execution. This integration
of human oversight into decision-making processes is critical to preventing
unintended harm and ensuring compliance with international humanitarian
law. Additionally, ongoing research into ethical-by-design architectures aims
to build ethical constraints directly into the system’s operational framework,
limiting the scope of actions that an autonomous system can take based on
predefined ethical guidelines.
Finally, the deployment of swarm intelligence in autonomous drones and
robots introduces challenges related to collective decision-making and dis-
tributed accountability. In swarm systems, decisions are often made collec-
tively by a distributed group of agents, with no single agent being responsible
for the final outcome. This creates significant ethical ambiguity in determining
accountability when swarm systems malfunction or cause harm. Solutions
such as distributed ledger technologies (DLT), including blockchain, have
been proposed to ensure that every decision made within the swarm is
recorded in a transparent and immutable way, providing a traceable log of
actions that can be audited for accountability purposes.
e2463722-34 P. RADANLIEV
Discussion
The introduction of fairness, transparency, and accountability into AI systems,
while crucial for ensuring ethical standards, introduces a significant financial
burden and operational complexity, especially in sectors where fast innovation
is a competitive necessity.
One of the primary economic costs arises from the increased complexity in
developing AI systems that adhere to ethical guidelines. Implementing fair-
ness-aware learning algorithms, such as demographic parity or equalized odds,
requires additional computational resources and extensive testing during the
training phase. These fairness constraints are not simply add-ons but require
a fundamental rethinking of the algorithmic design, particularly in cases where
performance optimization conflicts with fairness. For instance, in financial
services, ensuring that loan approval algorithms do not exhibit bias may
necessitate retraining models with diverse datasets and applying fairness con-
straints throughout the development cycle. This extended development pro-
cess incurs higher labor costs, requires greater infrastructure investment, and
often results in longer timeframes to achieve regulatory compliance.
Additionally, privacy-preserving techniques, such as differential privacy and
federated learning, add further complexity. Federated learning, which enables
model training across distributed datasets without centralizing sensitive data,
requires more sophisticated system architectures and secure communication
channels, increasing both the cost and technical difficulty of implementation.
Operationally, the impact of strict ethical guidelines is felt through the need
for ongoing compliance and continuous monitoring of AI systems. Ethical
frameworks such as the European Union’s AI Act mandate that high-risk AI
applications, particularly in fields like healthcare and criminal justice, undergo
continuous auditing to ensure ethical standards are maintained post-
deployment. These operational costs are amplified by the need to integrate
real-time fairness monitoring tools, such as AI Fairness 360, which check for
bias drift or decision-making anomalies as AI systems encounter new data.
These tools require continuous computational resources, infrastructure sup-
port, and personnel dedicated to auditing and model recalibration. For indus-
tries such as financial services, where AI systems are deployed in real-time
environments like high-frequency trading, maintaining fairness and compli-
ance adds layers of complexity to the operational workflow. This constant
need for recalibration can also result in downtime, during which systems must
be reevaluated and updated, leading to delays in decision-making processes
and potential disruptions to business continuity.
The financial impact of these ethical requirements also affects innovation
cycles and speed to market. In highly competitive sectors like autonomous
driving or AI-driven diagnostics, time-to-market is often crucial for gaining
a first-mover advantage. Companies that invest heavily in ethical compliance –
APPLIED ARTIFICIAL INTELLIGENCE e2463722-35
such as model transparency, fairness audits, and explainability – may experi-
ence delays in bringing products to market. For instance, the requirement to
integrate explainability mechanisms, such as SHAP (Shapley Additive
Explanations) or LIME (Local Interpretable Model-agnostic Explanations),
into AI models often necessitates additional development and testing phases.
This extends the overall project timeline and may place companies at
a competitive disadvantage against those who prioritize rapid deployment
over ethical oversight. The delay not only impacts short-term revenue but
also affects long-term strategic positioning, particularly in industries where
technological leadership is key to maintaining market share.
Beyond development and operational costs, legal compliance and regula-
tory risk are significant financial considerations for companies implementing
strict ethical guidelines. Regulatory frameworks like the GDPR and the
upcoming EU AI Act impose severe penalties for noncompliance, with fines
that can reach up to 4% of a company’s global revenue for violations of data
privacy and transparency requirements. To mitigate these risks, companies
often need to invest heavily in legal teams, external audits, and compliance
infrastructures. This introduces an additional cost layer as companies must
allocate resources not just for initial development but also for ongoing com-
pliance management. The cyclical nature of compliance – where systems must
be continuously updated, audited, and re-certified to meet evolving stan-
dards – creates long-term financial commitments that extend well beyond
the initial implementation of AI systems.
Despite these costs, emerging technologies offer potential solutions
that could mitigate some of the financial and operational burdens
associated with ethical AI. Automated machine learning (AutoML) sys-
tems are increasingly capable of incorporating fairness and transparency
checks into their development pipelines, reducing the need for manual
intervention and thus lowering labor costs. Additionally, distributed
ledger technologies (DLT), such as blockchain, can help track AI deci-
sions in a transparent and immutable way, thereby simplifying post-
deployment audits and reducing the cost of maintaining ethical stan-
dards. Nevertheless, while these technologies offer some relief, they
come with their own set of technical challenges and infrastructural
costs, which require additional investment and expertise to implement
effectively.
The implementation of strict ethical guidelines in AI development signifi-
cantly impacts economic and operational aspects of AI projects. While these
guidelines are crucial for ensuring fairness, transparency, and accountability,
they introduce substantial costs at every stage of the AI lifecycle, from devel-
opment through to post-deployment monitoring and compliance. Balancing
these ethical obligations with the need for innovation and market competi-
tiveness remains a challenge, particularly for companies operating in highly
e2463722-36 P. RADANLIEV
dynamic and competitive sectors. The evolving landscape of AI governance,
coupled with emerging cost-saving technologies, will be critical in determining
how companies navigate the financial and operational implications of ethical
AI development.
Conclusion
This study has undertaken an examination of the ethical imperatives sur-
rounding AI, particularly the principles of transparency, fairness, and privacy,
in the context of its prevalent influence across sectors such as healthcare,
finance, and communication. The deployment of AI technologies in these
domains brings with it profound ethical challenges that necessitate a strong
and inclusive framework to safeguard individual rights and societal interests.
Through a comparative analysis of international AI policy frameworks from
the European Union, the United States, and China, this research has clarified
the conflicting ethical priorities that shape AI governance globally.
This work clarifies the ethical principles of privacy, transparency, and fair-
ness, addressing regional challenges and interdependencies. By distinguishing
how these principles operate independently yet interactively across frame-
works, the paper offers a refined conceptual foundation necessary for global
governance. A primary contribution is the proposed set of integration criteria
(Interoperability, Normative Cohesion, Cultural Adaptability, and
Transparency of Process). These criteria provide a structured foundation for
aligning ethical principles across diverse international frameworks, supporting
cross-border AI compatibility while respecting region-specific values and
regulatory approaches. The graphical representations represent the individual
and corelated interdependencies and conflicts among frameworks, avoiding
oversimplification and enhancing analytical clarity. This provides a visual tool
for understanding ethical elationships in global AI governance.
The analysis reveals marked variations in how different regions balance the
demands of innovation against the ethical principles of privacy, fairness, and
accountability. While certain jurisdictions, such as the European Union,
emphasize stringent regulatory oversight and data protection, others, includ-
ing the United States, adopt a more flexible, innovation-centric approach.
These divergences underscore the complexities involved in striving for
a harmonized global standard for ethical AI governance. Nevertheless, this
study has articulated several strategic interventions to mitigate algorithmic
bias, including the deployment of fairness-aware algorithms, regular audits,
and the incorporation of diverse development teams. These interventions are
essential in fostering equitable and trustworthy AI systems.
Furthermore, this research has highlighted the critical need for sustained
international collaboration and dialogue to bridge the gaps in global AI ethics
frameworks. It is increasingly evident that no single jurisdiction can fully
APPLIED ARTIFICIAL INTELLIGENCE e2463722-37
address the multi-faceted ethical challenges posed by AI in isolation. Instead,
the path forward demands a concerted, cooperative effort that leverages shared
principles while respecting regional variations in regulatory and cultural
priorities.
This study advances the argument that ethical considerations must be
embedded at every stage of the AI development lifecycle, from inception
through to deployment and beyond. The recommendations herein aim to
inform policymakers, regulators, and AI developers, encouraging the pursuit
of AI systems that are not only innovative and technologically advanced but
also aligned with the highest ethical standards. As AI continues to evolve and
exert its transformative potential, the need for vigilance, adaptability, and
cross-border cooperation remains paramount in ensuring that these technol-
ogies serve the common good, promoting fairness, accountability, and trust in
their application.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Funding
The work was supported by the Engineering and Physical Sciences Research Council [EP/
S035362/1].
ORCID
Petar Radanliev http://orcid.org/0000-0001-5629-6857
Data Availability Statement
The datasets generated and analyzed during the current study are available from the corre-
sponding author upon reasonable request. Due to the sensitive nature of the data related to AI
ethics and privacy considerations, access to the data may be restricted. Specific details regard-
ing the data sources, including international AI policy frameworks from the EU, US, China,
Canada, Japan, India, and Australia, are documented within the study. All data shared will be
compliant with ethical guidelines and privacy standards as outlined in the General Data
Protection Regulation (GDPR) and other relevant data protection laws.
References
Aldoseri, A., K. N. Al-Khalifa, and A. M. Hamouda. 2023. Re-Thinking data strategy and
integration for artificial intelligence: Concepts, opportunities, and challenges. Applied
Sciences 13 (12):7082. doi: 10.3390/APP13127082 .
e2463722-38 P. RADANLIEV
Bécue, A., I. Praça, and J. Gama. 2021. Artificial intelligence, cyber-threats and industry 4.0:
Challenges and opportunities. Artificial Intelligence Review 54 (5):3849–86. doi: 10.1007/
s10462-020-09942-2 .
Bender, E. M., T. Gebru, A. McMillan-Major, and S. Shmitchell. 2021. On the dangers of
stochastic parrots: Can language models be too big? FAccT 2021 - Proceedings of the 2021
ACM Conference on Fairness, Accountability, and Transparency, 610–23. doi: 10.1145/
3442188.3445922 .
Binns, R. 2018. Fairness in machine learning: Lessons from political philosophy. Proceedings of
Machine Learning Research, vol. 81, 149–59, PMLR. https://proceedings.mlr.press/v81/
binns18a.html .
Bommasani, R., K. Klyman, D. Zhang, and P. Liang. 2023. Do foundation model providers
comply with the draft EU AI act? Center for Research on Foundation Models (CRFM):
Stanford Center for Research on Foundation Models.
Bostrom, N., and E. Yudkowsky. 2014. The ethics of artificial intelligence. The Cambridge
Handbook of Artificial Intelligence 316–34. doi: 10.1017/CBO9781139046855.020 .
Brynjolfsson, E., and A. Mcafee. 2014. The second machine age: Work, progress, and prosperity
in a time of brilliant technologies. In The second machine age: Work, progress, and prosper-
ity in a time of brilliant technologies. Worldwide: W.W. Norton & Company 978-0-393-
35064-7 https://wwnorton.com/books/the-second-machine-age/ .
de Bruin, B., and L. Floridi. 2017. The ethics of cloud computing. Science and Engineering
Ethics 23 (1):21–39. doi: 10.1007/s11948-016-9759-0 .
de Fine Licht, K., and J. de Fine Licht. 2020. Artificial intelligence, transparency, and public
decision-making. AI & Society 35 (4):917–26. doi: 10.1007/s00146-020-00960-w .
Du, M. 2023. Awesome-Fairness-in-AI. GitHub Repository. https://github.com/datamllab/awe
some-fairness-in-ai .
European Parliament. 2023. AI act: A step closer to the first rules on artificial intelligence | news |
European Parliament. https://www.europarl.europa.eu/news/en/press-room
/20230505IPR84904/ai-act-a-step-closer-to-the-first-rules-on-artificial-intelligence .
Evans, K. 2015. The second machine age: Work, progress, and prosperity in a time of brilliant
technologies by eric Brynjolfsson and Andrew McAfee. Journal of Business & Finance
Librarianship 20 (3):244–46. doi: 10.1080/08963568.2015.1044355 .
FACT SHEET: Biden-Harris Administration Announces New Actions to Promote Responsible
AI Innovation That Protects Americans’ Rights and Safety | The White House. 2023. https://
www.whitehouse.gov/briefing-room/statements-releases/2023/05/04/fact-sheet-biden-
harris-administration-announces-new-actions-to-promote-responsible-ai-innovation-that-
protects-americans-rights-and-safety/ .
Floridi, L., J. Cowls, M. Beltrametti, R. Chatila, P. Chazerand, V. Dignum, C. Luetge,
R. Madelin, U. Pagallo, F. Rossi, et al. 2018. AI4People—an ethical framework for a good
AI society: Opportunities, risks, principles, and recommendations. Minds and Machines
28 (4):689–707. doi: 10.1007/s11023-018-9482-5 .
GDPR. 2018. What is GDPR, the EU’s new data protection law? - Gdpr.Eu. https://gdpr.eu/
what-is-gdpr/ .
Helbing, D., B. S. Frey, G. Gigerenzer, E. Hafen, M. Hagner, Y. Hofstetter, J. Van Den Hoven,
R. V. Zicari, and A. Zwitter. 2018. Will democracy survive big data and artificial intelligence?
In Towards digital enlightenment: Essays on the dark and light sides of the digital revolu-
tion, 73–98. Springer International Publishing. doi: 10.1007/978-3-319-90869-4_7 .
HIPAA. 1996. Health insurance portability and accountability act of 1996 (HIPAA) | CDC.
https://www.cdc.gov/phlp/publications/topic/hipaa.html .
Hosny, A., C. Parmar, J. Quackenbush, L. H. Schwartz, and H. J. W. L. Aerts. 2018. Artificial
intelligence in radiology. Nature Rev Cancer 18 (8):500. doi: 10.1038/S41568-018-0016-5 .
APPLIED ARTIFICIAL INTELLIGENCE e2463722-39
IBM. 2018. AI fairness 360 – open source. Open Project. https://www.ibm.com/opensource/
open/projects/ai-fairness-360/ .
ICO. 2018. Information commissioner’s office (ICO): The UK GDPR. UK GDPR Guidance and
Resources. https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/lawful-
basis/a-guide-to-lawful-basis/lawful-basis-for-processing/consent/ .
Interim Measures for the Management of Generative Artificial Intelligence Services, Personal
Information Protection Law of the People’s Republic of China (PRC). 2023.
ISO. 2023. ISO/IEC DIS 42001 - information technology — artificial intelligence — management
system. https://www.iso.org/standard/81230.html .
Jobin, A., M. Ienca, and E. Vayena. 2019. The global landscape of AI ethics guidelines. Nature
Machine Intelligence 2019 1 (9):389–99. doi: 10.1038/s42256-019-0088-2 .
Li, L. 2017. China’s manufacturing locus in 2025: With a comparison of “Made-in-China 2025”
and “Industry 4.0”. Technological Forecasting & Social Change 135:66–74. doi: 10.1016/J.
TECHFORE.2017.05.028 .
Malhotra, Y. 2018. Cognitive computing for anticipatory risk analytics in intelligence, surveil-
lance, & reconnaissance (ISR): Model risk management in artificial intelligence & machine
learning (presentation slides). SSRN Electronic Journal. doi: 10.2139/ssrn.3111837 .
McCorduck, P., and C. Cfe. 2004. Machines who think: A personal inquiry into the history and
prospects of artificial intelligence. CRC Press. https://books.google.com/books?hl=en&lr=
&id=r2C1DwAAQBAJ&oi=fnd&pg=PP1&dq=Pamela+McCorduck+%22Machines+Who
+Think&ots=UnmXIiuRtM&sig=JAh90Eu07MGvjS5OgFq1CMy6-gc .
Meissner, G. 2020. Artificial intelligence: Consciousness and conscience. AI & Society
35 (1):225–35. doi: 10.1007/s00146-019-00880-4 .
MeitY. 2023. Artificial intelligence committees reports | Ministry of electronics and informa-
tion technology, Government of India. Artificial Intelligence Committees Report. https://
www.meity.gov.in/artificial-intelligence-committees-reports .
Mijwil, M. M., M. Aljanabi, and ChatGPT. 2023. Towards artificial intelligence-based cyberse-
curity: The practices and ChatGPT generated ways to combat cybercrime. Iraqi Journal for
Computer Science and Mathematics 4 (1):65–70. doi: 10.52866/IJCSM.2023.01.01.0019 .
Mittelstadt, B. 2019. Principles alone cannot guarantee ethical AI. Nature Machine Intelligence
1 (11):501–07. doi: 10.1038/s42256-019-0114-4 .
Mozumder, M. A. I., M. M. Sheeraz, A. Athar, S. Aich, and H.-C. Kim. 2022. Overview:
Technology roadmap of the future trend of metaverse based on IoT, blockchain, AI
technique, and medical domain metaverse activity. International Conference on Advanced
Communication Technology (ICACT) 256–61. doi: 10.23919/ICACT53585.2022.9728808 .
NAIAC. 2024. AI safety: National AI advisory committee. https://ai.gov/wp-content/uploads/
2024/06/FINDINGS-RECOMMENDATIONS_AI-Safety.pdf .
NIST. 2023a. AI risk management framework | NIST. National Institute of Standards and
Technology. https://www.nist.gov/itl/ai-risk-management-framework .
NIST. 2023b. Artificial intelligence | NIST. https://www.nist.gov/artificial-intelligence .
NIST. 2024a. AI risk management framework | NIST.
NIST. 2024b. AI standards | NIST. https://www.nist.gov/artificial-intelligence/ai-standards .
NIST. 2024c. Department of commerce announces new guidance, tools 270 days following
president Biden’s executive order on AI | NIST. https://www.nist.gov/news-events/news/
2024/07/department-commerce-announces-new-guidance-tools-270-days-following .
Office for Artificial Intelligence and Department for Science, Innovation & Technology. 2023.
A pro-innovation approach to AI regulation 978-1-5286-4009-1 (London: Crown copyright).
Partnership on AI. 2023. Partnership on AI and the ethical AI framework for social good. https://
partnershiponai.org/ .
e2463722-40 P. RADANLIEV
Provisions on the Administration of Deep Synthesis Internet Information Services, Personal
Information Protection Law of the People’s Republic of China (PRC). 2022.
Roberts, H., J. Cowls, J. Morley, M. Taddeo, V. Wang, and L. Floridi. 2021. The Chinese
approach to artificial intelligence: An analysis of policy, ethics, and regulation. AI & Society
36 (1):59–77. doi: 10.1007/s00146-020-00992-2 .
Shu, Y., J. Zhang, and H. Yu. 2021. Fairness in design: A tool for guidance in ethical artificial
intelligence design. Lecture Notes in Computer Science (Including Subseries Lecture Notes in
Artificial Intelligence and Lecture Notes in Bioinformatics) 12774:500–10. doi: 10.1007/978-
3-030-77626-8_34 .
Singer, P. W. 2009. Wired for war: The robotics revolution and conflict in the twenty-first
century, 499. https://books.google.com/books/about/Wired_for_War.html?id=
AJuowQmtbU4C .
The State Council People Republic of China. 2017. Made in China 2025; the state council people
Republic of China. http://english.gov.cn/2016special/madeinchina2025/ .
Tabassi, E. 2023. AI risk management framework | NIST. doi: 10.6028/NIST.AI.100-1 .
Turilli, M., and L. Floridi. 2009. The ethics of information transparency. Ethics and Information
Technology 11 (2):105–12. doi: 10.1007/s10676-009-9187-9 .
UNESCO. 2023. Recommendation on the ethics of artificial intelligence | UNESCO. https://www.
unesco.org/en/articles/recommendation-ethics-artificial-intelligence .
Wachter, S., B. Mittelstadt, and C. Russell. 2023. Health care bias is dangerous. But so are
fairness’ algorithms | WIRED. Wired. https://www.wired.com/story/bias-statistics-artificial-
intelligence-healthcare/ .
Yu, K. H., A. L. Beam, and I. S. Kohane. 2018. Artificial intelligence in healthcare. Nature
Biomedical Engineering 2 (10):719–31. doi: 10.1038/S41551-018-0305-Z.
APPLIED ARTIFICIAL INTELLIGENCE e2463722-41
