Exploring the significance and implementation of trustworthiness in artificial intelligence (AI) within the industrial context forms the core theme of this study, highlighting the crucial role of AI's expanding influence in key sectors. It emphasizes key trustworthiness dimensions including robustness, transparency, and fairness, essential for user trust. The work discusses the challenges in implementing these principles, particularly in ethical integration and public trust maintenance. A focus is placed on the EU's AI Act, which introduces a risk-based regulatory framework categorizing AI systems into various risk levels with corresponding regulations. Additionally, the paper examines the relationship between ethics and AI legislation, noting the influence of ethical guidelines on regulatory practices. It then proposes a Trustworthy System Framework for Zero Defect Manufacturing in the industry, incorporating "Trustworthy Pillars", compliance with regulations, robust technical infrastructure, human interaction considerations, process integrity, and operational stability. This framework is designed to enhance the trustworthiness of AI systems in an industrial setting.
2. The AI Act: What Do the Latest Developments Mean for
AIpowered Industry?
The rapid progress of AI-driven technologies has led to the urgent need for an adequate and
up-todate legal framework. While aiming to boost research and industrial capacity, ensuring safety and
protection of fundamental rights is of supreme importance [
The process of drafting has taken a couple of years, and the EU has worked on different aspects
and identified the most vital standards that must be covered by the legislative response. The final
steps in this regard were taken in 2023. In June 2023, the European Parliament released its official
position, thereby announcing the main requirements to be included: ban of biometric surveillance,
emotion recognition, predictive policing, the need of disclosure that content was AI-generated, as
well as the high-risk effect of the involvement of AI systems in elections [
In December 2023, The Members of the European Parliament (hereinafter “MEPs”) reached a
political deal with the Council on a Bill to ensure AI in Europe is safe, respects fundamental rights
and democracy, while businesses can thrive and expand. The concluded negotiations are a
provisional agreement on the AI Act, with the overall unifying goal being to ensure that
fundamental rights, democracy, the rule of law and environmental sustainability are protected from
high-risk AI, while boosting innovation and making Europe a leader in the field. The rules establish
obligations for AI based on its potential risks and level of impact [
Over the next few weeks, representatives from the EU institutions shall keep polishing any
outstanding technical aspects. In the first half of 2024, the final text shall be presented for approval
to the European Parliament and the Council. The approved version shall be published in the Official
Journal after it has been translated into the EU's official languages. The implementation period shall
then commence twenty days after the publication of the AI Act [
It has been agreed that the goal of the EP is to ensure that created and used AI systems are “safe,
transparent, traceable, non-discriminatory and environmentally friendly”. The prevention of
harmful outcomes has also been addressed when it comes to human monitoring. A risk-assessment
methodology has been used by the Parliament when approaching this legislation. In other words, the
level of potential risks must be addressed and based on that, obligations may be imposed on
providers or users. The following paragraphs are dedicated to the types of identified risks [
Unacceptable Risk: The AI systems that fall under this category are considered as people
threatening and are going to be prohibited. They could include: “Cognitive behavioural
manipulation of people or specific vulnerable groups: for example, voice-activated toys that
encourage dangerous behaviour in children; Social scoring: classifying people based on behaviour,
socio-economic status or personal characteristics; Real-time and remote biometric identification
systems, such as facial recognition”. Potential exceptions could be introduced such as related to
crime prevention [
High Risk: The high-risk AI systems are such that can be a treat to safety or fundamental rights.
Parties that are creating them should comply with regulations that demand thorough testing,
appropriate documentation and a responsibility structure that envisages human oversight. They
shall have to be analysed and evaluated prior to being allowed to be distributed, as well as after that.
According to the AI Act, they shall be divided into 2 main categories [
Limited Risk: Only minimal transparency requirements shall be imposed for this category, with
the aim being to give the opportunity to users to make informed decisions. The system must make it
clear that users are interacting with AI [
Minimal or No Risk: Most of the current AI systems in the EU fall under this category,
including applications such as AI-enabled video games or spam filters [
Generative AI shall have to implement transparency requirements [
Apart from assessing the danger posed by AI systems, safeguarding the rights of citizens is also a
top priority. Nonetheless, research on AI-based systems may be conducted under open-source
licenses to further encourage innovation. Before the system is released to the market, it is integrated
into an open test environment for public testing. In addition, the public is urged to make complaints
as appropriate and to be aware of choices made using high-risk AI systems that may jeopardize their
rights [
Constituting e set of moral principles, ethics assist in discerning between right and wrong, thereby
serving as guidelines for best practice. Engaged with the study of optimizing AI’s beneficial impact
while reducing risks and adverse outcomes, AI ethics is thus a multidisciplinary field. Ethics has a
motivating role during regulatory and legislative processes and oftentimes serves as an inspiration
for the creation of new regulatory instruments, for revisions of existing legislation, or for abolishing
such. Yet, what ethics is not capable of achieving, but regulation does, is codifying and enforcing
ethically desirable behavior [
An ongoing debate has been taking place in respect of having ethical guidelines and/or principles
navigating AI development instead of actual AI regulation. The two colliding views being that, on
the one hand, ethics-driven self-regulation can take the place of external regulation, whilst, on the
other, it is recalled that the principles do not provide clarity as to how it may be ascertained that a
company follows AI ethics principles. To avoid misunderstanding the role of ethical governance, it
must be noted that ethics is not intended to replace regulation. It is a way to conclude what kind of
regulation is needed. Therefore, to safeguard that AI ethics possess the ability to uncover ethical
issues in a timely manner, thereby serving as a step towards creating appropriate legislation, ethics
must be able to accompany the development of AI [
With ongoing developments both on national and international fronts, the rather infant legislative
framework governing AI is inevitably sure to come. As oftentimes, legislation lags innovation,
particularly in the context of AI, proactive measures may still be applied through the help of the
principles of ethical use of AI. For example, the Ethics Guidelines for Trustworthy AI (April 2019) [
One of the key components of this framework is the identification and incorporation of what are known as the "Trustworthy Pillars." These pillars are derived from the main characteristics that have been identified through the literature review. Moreover, these pillars align with the different facets of trustworthiness as described by [16]. This standard creates emphasis on the universality of these trustworthiness facets, asserting that they are relevant not only to all systems but also to the core services within the industrial sector as well.
It is important to note that the environment in which a system operates plays a crucial and significant role in determining its overall trustworthiness. This encompasses the system's compliance with various external regulations, including those that are mandated by the EU and other countries where the industrial solutions are deployed. To further elaborate, regulations are defined by the Organisation for Economic Co-operation and Development (OECD) and the Centre for Co-operation with European Economies in Transition [17] as the imposition of rules that are enforced by the government, with penalties being imposed for non-compliance. These regulations, along with the internal and external standards that are in place, are integral components of the Trustworthy System Framework (TSFr) that has been specifically designed for the industrial sector (Figure 1).
The integration of core services into a multi-layered framework, as suggested by [18], efficiently structures software development by segregating functionalities into distinct layers such as advisory, analytical, and data layers. This architecture enhances modularity, facilitating focused development within each layer while maintaining system-wide cohesion. A generic adaptation of this model includes additional layers like edge, storage, and visualization, interconnected through specialized services for data integration, transformation, and secure communication. Key to this framework is the incorporation of security mechanisms and workload optimization solutions, such as blockchain for traceability and AI for edge computing efficiency. This design approach not only simplifies complex software system development but also boosts scalability, flexibility, and security, offering a robust foundation for building dependable software solutions.
Another critical factor which has significant influence on the trustworthiness of a system is its technical infrastructure. The robustness and resilience of the infrastructure against various vulnerabilities, such as attacks, errors, or faults, are vital in instilling trust in the system. To enhance trustworthiness, a multitude of state-of-the-art technologies and methodologies are proposed, which cover a wide range of areas including the development of a comprehensive Data Quality Strategy, the establishment of Data Trustiness and Traceability, the implementation of Data Trusted Communication and Distribution, the assurance of Data Security, and the efficient management of Data Storage and Use.
Furthermore, it is essential to recognize the significance of human factors within the context of semi- automatized industrial solutions. As [19] points out, human interaction with systems poses a potential risk factor that can impact the overall trustworthiness of the system. Therefore, it is of utmost importance to have a thorough understanding of and control over human interaction using conceptual models and methods. Processes, which encompass all hardware and software-related procedures that require human interaction, are identified as crucial elements within the Trustworthy System Framework (TSFr) for the industrial sector. It is important to acknowledge that any flaws or potential faults that may arise within these processes directly impact the overall trustworthiness of the system. Additionally, the use of open-source software libraries in industrial solutions is addressed within the framework. While open-source libraries offer certain benefits, they are often met with skepticism when it comes to trustworthiness. Consequently, a strict quality assurance procedure is mandated for all open-source libraries that are utilized.
The framework also acknowledges the importance of a system's operational integrity in the face of various external and internal events, which range from accidents and attacks to natural disruptions and system failures. In conclusion, the framework incorporates the concept of core services, as proposed by [18], in their Framework for Trusted Software Development. This concept encompasses different layers within the system, such as the advisory layer, analytical layer, and data layer, each of which has its own set of functionalities. These layers collectively form the structure of the entire industrial platform, ensuring interconnectedness and secure communication between various solutions, thereby reinforcing the system's overall trustworthiness.
The author(s) have not employed any Generative AI tools. [15] Thilo Hagendorff, “A Virtue‑Based Framework to Support Putting AI Ethics into Practice” (21
June 2022) , available at: <https://link.springer.com/article/10.1007/s13347-022-00553-z>. [16] British Standards Institution (2018). Information technology - Systems trustworthiness - Part 1:
Governance and management specification (BS 10754-1:2018). [17] OECD; Centre for Co-operation with European Economies in Transition. (1993). Glossary of industrial organisation economics and competition law. OECD. [18] Bose, R. P. J. C., Singi, K., Kaulgud, V., Phokela, K. K., & Podder, S. (2019). Framework for Trustworthy Software Development. In 2019 34th IEEE/ACM International Conference on Automated Software Engineering Workshop (ASEW) (pp. 45–48). IEEE. https://doi.org/10.1109/ASEW.2019.00027. [19] Schneider, F. B. (Ed.). (1999). Trust in Cyberspace. The National Academies Press. https://doi.org/10.17226/6161.