Artificial Intelligence (AI) is a fast-changing technology that is having a profound impact on our society, from education to industry. Its applications cover a wide range of areas, such as medicine, military, engineering and research. The emergence of AI and Generative AI have significant potential to transform society, but they also raise concerns about transparency, privacy, ownership, fair use, reliability, and ethical considerations. The Generative AI adds complexity to the existing problems of AI due to its ability to create machine-generated data that is barely distinguishable from human-generated data. Bringing to the forefront the issue of responsible and fair use of AI. The security, safety and privacy implications are enormous, and the risks associated with inappropriate use of these technologies are real. Although some governments, such as the European Union and the United States, have begun to address the problem with recommendations and proposed regulations, it is probably not enough. Regulatory compliance should be seen as a starting point in a continuous process of improving the ethical procedures and privacy risk assessment of AI systems. The need to have a baseline to manage the process of creating an AI system even from an ethics and privacy perspective becomes progressively more important In this study, we discuss the ethical implications of these advances and propose a conceptual framework for the responsible, fair, and safe use of AI.
science and technology that has the potential to revolutionize various sectors of industry and society. With its ability to process vast amounts of data, generate insights, and support decision-making, AI has emerged as an important part of many organizations’ processes. However, concerns about the impact of AI on society, particularly from an ethical perspective, have increased as its use has grown. From self-driving cars to virtual assistants, the applications of AI are endless as the quality and performance of AI techniques and methods continue to improve. ative AI is a subset of AI that uses Machine Learning (ML) plications of AI and increases the dangers it poses. Gener- of view. algorithms to generate new content based on existing data. It makes it possible to create content that appears as new and original, but is the result of generating statistics based on training data sets.
Generative AI raises new ethical challenges and a nized by CINI, May 29-30, 2024, Naples, Italy ∗Corresponding author. †These authors contributed equally. (R. D. Fazio) © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License
CEUR
ceur-ws.org whole new set of emerging issues because of the dificulty in separating human-generated content from machinegenerated content.
It becomes crucial a fair use of AI in any field of application, first and foremost in sensitive fields such as medical, military, and engineering, where the human decision-making component is of primary importance, but also in research and education where fair use of AI becomes critical to the informed growth of students with critical thinking and quality research. With the rapid developments in machine learning and generative AI models, the newborn of more powerful Large Language
ethical challenges associated with the use of generative AI. The technology has the potential to revolutionize various sectors of society. However, it also raises concerns about job displacement, transparency, privacy, ownership, inequality, and reliability. To ensure that, the benefits of generative AI are maximized while its risks are minimized, the development of responsible and ethical frameworks for its use will be critical.
3 focuses on the ethical implications and issues of AI. and related to ethical and legal aspects. Observing the
Section 4 presents the conceptual framework. Finally, phenomenon of the outstanding popularity of these kinds
Section 5 presents the conclusion and future research of systems and tools among common users, it brings back
directions. to mind the efects of Web 2.0, with the introduction of
User Generated Contents (UGC)[8], where people were
enabled to write everything almost everywhere. A
dele2. AI and Generative AI terious phenomenon deriving from the exceeding
democbackground racy of the web still remains represented by the fake
news unconditioned spreading[9], as discussed in the
In the last few years, Artificial Intelligence Generated studies proposed in[
Whether GAI represents a significantly challenging training. [19] issue for researchers, involved in understanding and im- In GAI, the amount of data used to train the model proving the representation of the knowledge that is be- is enormous. Often this data has been collected from hind, GAI-based systems are also carriers of not trivial the Internet and using diferent and heterogeneous data implications, as the ones represented by social impact sources. Unlike a traditional AI model, which is used consent of the people involved, possibly violating their privacy. • Identity Theft: by spreading false information, misleading content, or malicious messages, fake content can be used to impersonate individuals and cause significant damage to their reputations and privacy, as well as organize financial fraud. • Revenge Porn: fake content can be used to create not real videos or images that show people in compromising situations, damaging their privacy and reputation. In the most serious cases, money is solicited for extensive purposes. • Misinformation or Disinformation: fake content can have a significant impact on public opinion, trust, and decision-making by spreading false information or propaganda. This misinformation can also have a serious impact on society. It can lead to social unrest, political instability, and other negative consequences. for prediction or classification purpose, the generative model is used to create new content. Bias in training data increases, in this case a specific ethical issue, perpetuating a potential social bias in the newly generated content.
In addition, researchers working in this area face another ethical dilemma: if the data contain biases that reflect society, is it correct to work to mitigate these bi- It is important to emphasize that privacy issues arise ases? If so, how? ifrst in AI processes. In fact, there are significant privacy
Certainly, it must be done with the utmost care, be- issues at the data collection stage, because in this stage cause biased AI systems could potentially exacerbate is where sensitive information is collected and stored, existing societal inequalities. They could perpetuate prej- making it vulnerable to potential security breaches and udice or reinforce stereotypes. They could also produce unauthorized access. [22], [23] disparate outcomes for groups based on factors like race, Finally, it is also interesting for this discussion to mengender, or socioeconomic status, leading to further in- tion the problem of copyrighting the content on which equality and social unrest. There is a real risk of perpet- AI systems, especially GAI systems, are trained. Often uating harmful stereotypes and possibly even distorting the source of this data is not really known. beliefs [20]. GAI systems can use, process, and generate content
Strictly related to the bias issue, in the special mode without explicit consent, potentially violating the privacy with generative content creation, there is the problem of of individuals and organizations. misleading information and fake news generation. The considerations presented here on the ethical risks
The ability of LLMs to create information that is not associated with AI and the perils that arise from them, present in their original data, known as hallucination of depend in great part on unaccountable or unfair use of LLMs or more technically called emerging feature [21], AI, both by the creators of AI systems and by the end introduces the problem of creating misleading text, which users of such technologies. could easily become fake news. In the field of text generation, there are a lot of use
Moreover, recent developments in GAI allow not only cases where LLMs can help and improve the regular actext, but also images (figure 1), video, and audio to be tivities of students and researchers if it is used in a fair created, enabling non-technical people to efectively use way. these techniques through simple applications. GAI systems, such as ChatGPT, could be leveraged for
It is clear that illegal use of these technologies has led students to get ideas or insights on specific topics. If you to attempts at fraud and extortion, and can also lead to know well the idea that you want to write, then the GAI major legal and social problems. The illegal use of these could help you to write it without grammar mistakes, techniques to create images and videos that substitute especially if you are writing in non-native language. the face or other physical characteristics of one person This could greatly benefit non-native speakers, even for those of another. Because of their potential to cre- in academia, in a sort of democratization of the dissemiate believable and deceptive content that can be used nation of scientific thought, without having to resort to to spread misinformation or damage the reputation of expensive language revision services. individuals. The most relevant privacy issues include: On the other hand, an unfair and unethical use of this technology by students and researchers raises a very important ethical and legal problem related to authorship. • Privacy Violation: fake content can be used to
manipulate existing videos or images without the
The need to know whether a piece of content is humangenerated or machine-generated is becoming more relevant and critical.
ernments of various countries around the world have not stood still.
The United States has responded with the AI Bill of Right [24], which is not a regulation but only a white paper of recommendation from the White House Ofice of Science and Technology Policy. It outlines the main principles to be followed to pursue ethical issues in AI. It is a guideline for designing and deploying AI systems that respect human rights, enhance fairness, and protect personal privacy.
The European Commission has gone further with the EU AI Act, a “Proposal for a Regulation of the European Parliament and of the Council laying down harmonized rules on artificial intelligence and amending certain Union legislative acts” [25]. It is a fully-fledged legislative proposal that aims to address the risks associated with artificial intelligence systems.
The AI Act intends to ensure that AI systems are trustworthy, reliable, and beneficial to individuals and society.
Furthermore, once again the European Commission enacted the General Data Protection Regulation (GDPR) [26], which although not closely related to artificial intelligence, protects the privacy rights of European citizens, with particular emphasis on the automatic gathering and processing of personal data.
These documents provide a solid foundation for the development of a conceptual framework to assist researchers and companies in developing and deploying AI systems that are not only compliant but also address and attempt to solve AI ethical issues.
The four pillars on which the proposed framework is built are: systems. These practices should be organized with clear guidelines to avoid any misunderstanding or abuse of AI techniques.
Finally, a regular and cyclical risk assessment process specific to AI systems is required to promptly identify, evaluate, and prioritize potential risks associated with the development of AI systems.
Generative AI is all about creating artificial data that looks like the real thing. This super-realistic data can be a game-changer in many fields, from video games to medicine and finance, until the arts. The resulting production of GAI is sometimes referred to as “fake data”, to evidence that the contents were generated by an automatic process performed by a machine and not by a human being. GAI enables to generate fake but realistic images, to write new text, compose music, and even build chatbots that seem like chatting with real people. Besides the research eforts to improve the quality of the AI production, several ethical, legal and security issues need to be addressed.
It is apparent to need to address these issues systematically and beyond mere regulatory compliance. The development of a conceptual framework to address these issues should be a good starting point.
Future work will include improving the framework and exploring ways to make it more practical, including measures of the performance of ethical and responsible use of AI and GAI.
Moreover, we plan an in-depth look at the topic of detecting human user-generated texts from texts generated by a GAI, exploring existing techniques and towards new ones. Finally, we will continue research in the area of XAI (whose exploration began in[27] and [28]), which also extends to GAI, in order to improve the transparency of AI systems.