ChatGPT – another step towards the digital era or a threat to fundamental rights and freedoms?

Since we are living in the digital age, AI systems and scientific tools are omnipresent in many areas of private and public life. In healthcare, robotic surgery, cardiac ultrasound tools, clinical diagnosis based on machinelearning systems are some of the elements which demonstrate the rise of AI technology. The use of machine learning models to search medical data and uncover insights to help improve health outcomes and patient experiences is occurring almost on a daily basis. In e-commerce, chatbots are used by a large number of companies providing a big variety of services, as it has been already exposed. As Augusto (2021) and Kirwan and Zhiyong (2020) point out through their study “smart cities are also a representative example of AI technology.” The continued growth of urbanisation presents new challenges. According to a survey conducted by the United Nation’s Department of Economic and Social Affairs on 2018, 55% of the world’s population reside in urban areas. This is expected to rise to 68% by 2050(Department of Economic and Social Affairs of United Nations, 2018).Smart Cities are part of the solution to the growing challenges of urbanization which demand sustainable development and improvement of quality of life. Smart cities can use artificial intelligence to see their effect on the local environment, global warming, as well as the pollution level. Using AI and machine learning within pollution control and energy consumption, allows authorities and cities to make well informed decisions that are best for the environment.

The aforementioned references are just indicative illustrations of the ongoing and daily use of Artificial Intelligence in our lives. At EU level, ΑΙ is the most or at least among the most representative actions of implementation of the EU digital strategy. It reflects the digitalization which penetrates our society in all sectors of private and public life and puts a clear focus on data, technology, and infrastructure.

In addition, AI plays a major role in shaping Europe’s digital future. It constitutes one of the most important actions in order to empower people with a new generation of technologies and create a fair and competitive environment for people and businesses. The EU’s approach to artificial intelligence centres on excellence and trust, aiming to boost research and industrial capacity and ensure fundamental rights(EU digital agenda on Artificial Intelligence,2019).

The European approach to artificial intelligence (AI) will help build a resilient Europe for the Digital Decade where people and businesses can enjoy the benefits of AI. It focuses on 2 areas: excellence in AI and trustworthy AI. The European approach to AI will ensure that any AI improvements are based on rules that safeguard the functioning of markets and the public sector, and people’s safety and fundamental rights.

However, there are many legal and ethical concerns regarding AI, such as: can can AI substitute or replace human factor in public and private life? Is AI always trustworthy and can lead to safe conclusions?Will AI in the form of ChatGpt and similar ‘largelanguage models’ contribute to the further easier spreading of misinformation and falsehoods, thus endangering the very fundamentals of democratic societies? Ethical considerations and limitations are societal norms and protect society from transgressions, could we expect those Artificial Intelligence tools to be able to respond to ethical standards?Those questions describe the context of the present paper which aims to present the nature of ChatGPT, an innovative viral chatbot that within a minimal time of its existence, dominates the world of Artificial Intelligence and raise a series of concerns. Consequently, at first level, this study will focus on the legal regime which governs the use of ChatGPT, after giving a short demonstration of its function and services (Section I).Secondly, a critical approach will be pursued focusing on special issues regarding this new AI tool, such as its ethical dimension, its impact on education and on the exercise of fundamental human rights, especially of those of privacy and freedom of expression (Section II). Following that intense analysis the paper aims to lead to fruitful conclusions with the ultimate goal to enhance the establishment of a powerful, safe and trustful digital environment.

2.    Section I: the conformity of ChatGPT with the European Law

• I.    Describing the function and services of ChatGPT

Before proceeding to the legal framework ruling the use of ChatGPT it is necessary to provide a short presentation of its services. Therefore, ChatGPT is a large language model chatbot developed by OpenAI based on GPT-3.5. It has a remarkable ability to interact in conversational dialogue form and provide responses that can appear surprisingly human. It was created in November 2022 by San Francisco-based artificial intelligence company OpenAI which is specialized in developing Large Language Models (AI learning machine tools) (Montti, 2022). As it is explicitly announced through the official website of the company, (ChatGPT, 2022) “interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.” Despite its short history, this AI tool knows a tremendous and massive expansion since, according to the recent reports, jumped to a million users just five days after its founding in November 2022. This number is incredible in comparison to other popular online platforms which enjoyed such attraction in a greater time, as it is clearly demonstrated by the research conducted by Ahmed (2023) and Buchholz (2023). All those elements verify that ChatGPT constitutes a milestone in the recent AI era, knowing a huge and ongoing acceleration in its development.

That AI tools provides multiple services in various areas, such as in commerce, online shopping, employment and scientific domains. ChatGPT can also be used to create interactive storytelling experiences, allowing users to explore and learn from virtual worlds. Some use cases for ChatGPT include:

• •    Generating responses in a chatbot or virtual assistant, to provide more natural and engaging interactions with users;

• •    Brainstorming content ideas on keywords or topics;

• •    Creating personalized communication, such as email responses or product recommendations (Marr, 2022).

• II.    Legal governance of ChatGPT.

Being an inherent part of AI, it becomes apparent that the legal rules which govern Artificial Intelligence are applying to that new AI model. At this point, it should be noticed that, for the time being, there is not a uniform and binding regulatory framework at the field, something which poses serious concerns regarding the protection of privacy and security as well as of other fundamental rights and freedoms, as Kouroupis and Serotila (2022) thogoughly analyze in their latest book (pp. 97-128).

On 21 April 2021, the European Commission (“Commission”) adopted a proposal for a “Regulation laying down harmonized rules on Artificial Intelligence” (“AI Regulation”), which sets out how AI systems and their outputs can be introduced to and used in the European Union. The draft AI Regulation is accompanied by a proposal for a new Regulation on Machinery Products, which focuses on the safe integration of the AI system into machinery, as well as a new Coordinated Plan on AI outlining the necessary policy changes and investment at Member State level to strengthen the EU’s leading position in trustworthy AI.1

According to the official pressrelease published the European Commission regarding the newrules for Artificial Intelligence (European Commission, 2021). Thereareadoptedriskcategories based on the intended purpose of the AI system, in line with the existing EU product safety legislation. Those categories are the following: unacceptable, high, limited and minimal risk. The criteria for this classification include the extent of the use of the AI application and its intended purpose, the number of potentially affected persons, the dependency on the outcome and the irreversibility of harms, as well as the extent to which existing Union legislation provides for effective measures to prevent or substantially minimise those risks.

The ChatGPT corresponds to the second level of risk, that of high-risk. Since it interacts in a conversational way, it develops a “human behavior”, having a strong impact on fundamental rights, such as those of freedom of expression and right to education. Therefore, requirements such as those of transparency, security and the provision of information to users, should be fulfilled.

Furthermore, it is crucial to underline that besides the AI rules, the General Data Protection Regulation shall be also respected. Since this learning language model operates on the base of a record of personal information provided by individuals, all the relevant provisions and principles regarding the lawfulness of processing personal data are in force. At this point a crucial clarification should be added: The only data protection regulation OpenAI mentions on its privacy policy page is the California Consumer Privacy Act (CCPA), which applies in the state where the company is based. For instance, the company states that California residents have “the rights to know what Personal Information we have collected and how we have used and disclosed that Personal Information; the right to request deletion of your Personal Information; and the right to be free from discrimination relating to the exercise of any of your privacy right.” Consequently, it appears not obvious or certain whether ChatGPT is GDPR compliant (Poireault, 2023).

In addition, it is of primary interest to underline that Artificial Intelligence poses new challenges because it enables machines to “learn” and to take and implement decisions without human intervention. Yet, decisions taken by algorithms could result from data that is incomplete and therefore not reliable, they may be tampered with by cyber-attackers, or they may be biased or simply mistaken. The ChatGPT is a representative example of the above declaration. Unreflectively applying the technology as it develops would therefore lead to problematic outcomes as well as reluctance by citizens to accept or use it. In that terms, AI technology, thus the ChatGPT, should be developed in a way that puts ethic values at the summit of all plans and goals.

The need for ethics guidelines led to the establishment by the Commission of a high-level expert group on AI representing a wide range of stakeholders (Shaping Europe’s digital future). The group is responsible on drafting AI ethics guidelines as well as preparing a set of recommendations for broader AI policy. The AI high-level expert group published a first draft of the ethics guidelines in December 2018. Following a stakeholder consultation and meetings with representatives from Member States, the AI expert group has delivered a revised document to the Commission in March 2019. In their feedback so far, stakeholders overall have welcomed the practical nature of the guidelines and the concrete guidance they offer to developers, suppliers and users of AI on how to ensure trustworthiness.

Guidelines for trustworthy AI drafted by the AI high-level expert group : The guidelines postulate that in order to achieve ‘trustworthy AI’, three components are necessary: (1) it should comply with the law, (2) it should fulfil ethical principles and (3) it should be robust. the guidelines identify seven key requirements that AI applications should respect to be considered trustworthy. The seven key requirements are the following (Ethics Guidelines for Trustworthy AI:

Human agency and oversight , Technical robustness and safety , Privacy and data governance , Transparency , Diversity, non-discrimination and fairness , Societal and environmental well-being and Accountability .

• III.    A holistic approach of the ChatGPT regime

After having demonstrated the legal framework as well as the ethical principles which (should) penetrate the use of ChatGPT, there will be attempted a critical approach regarding the applicability of the exposed rules on that new AI tool. For purposes of better consideration and evaluation of the subject matter, the ethical dimensions will be thoroughly studied and analyzed during the second chapter of the paper.

Therefore, the concern that arises is whether a compliance of ChatGPT with AI and GDPR rules can be achieved. The greatest interest is focused on the protection of subject’s rights as well as of the respect of the principles relating to processing of personal data (General Data Protection Regulation, Article 5). Hence, it follows a critical analysis of the aforementioned statement, based on the conformity of ChatGPT with several articles of the GDPR:

• Article 5:    Principles relating to processing of personal data .

  – Principle of transparency: it appears dubious the extent of the respect of this principle since the type of the personal information collected either is not defined or it is quite vague, covering a wide range of personal data such as social, communication and technical information.

• -    Principles of purpose limitation and data minimisation: since this language machine learning model processes a huge sort and different categories of personal data for any circumstance needed, it is not certain how those two principles are respected.

• –    Storage limitation: there is not a defined and clear provision regarding this issue.

• –    Integrity and confidentiality: it is explicitly mentioned that “ in certain circumstances we may share your Personal Information with third parties without further notice to you, unless required by the law, including without limitation in the situations ” such as vendors and Service Providers, Business Transfers etc. Consequently,ChatGPT has the ability to share personal data – from its training datasets – with its users. However, this functionality probably breaches european data protection laws since the undefined procession of personal data does not ensure appropriate security, leading indirectly (or even directly) to an unauthorised or unlawful processing.

• Article 6:    Conditions for consent.

3.    Section II: Ethical considerations of ChatGPT: challenges for journalism

In the last section of its privacy policy, it is mentioned the following: “ By continuing to use our Service or providing us with Personal Information after we have posted an updated Privacy Policy, or notified you by other means, you consent to the revised Privacy Policy ”. However, the GDPR requires consent to be opt-in. It defines consent as “ freely given, specific, informed and unambiguous ” given by a “ clear affirmative action ”. It is not acceptable to assign consent through the data subject’s silence. Therefore, there is clear violation of the article 6 of the GDPR.

Articles 12-23: Rights of the data subject.

Regarding that section, it becomes apparent that many rights may be violated since, according to the privacy policy of the AI tool, in many cases there is an obscure and unlawful processing of data rights. Indicatively, the right to be forgotten which is predicted under article 17 of the GDPR, is frequently violated due to the performance of ChatGPT. More specifically, this AI model tool does not collect any new information from the internet itself. Rather, it utilizes the data it already knows to generate responses. A lot of personal data from people who are often talked about on the web is included in its pre-existing dataset. In addition, it uses a machine learning method called transfer learning, which involves a model that is first trained on a data-rich task and then fine-tuned to a separate task. In this case, the model is pre-trained on a massive collection of words taken from the web and then tuned for the particular activity, according to Hilemann (2023). Consequently, in case of exercise the right to be forgotten, the AI learning machine tool should comprehend what data is used to create the archive of responses. However, this is neither certain nor an easy process since there is a procession of vague categories of personal data. In addition, in case of positive response to the right to be forgotten, the information requested shall be deleted. That would harm the efficient performance of the AI tool; Hence, the data processor would most probably deny the fulfillment of the right.

Researchers have long debated the importance of the freedoms of information and communication in the establishing and furthering of any democracy. Reflected in the journalistic work and the media are the potential, the dynamics and the shortcomings of that given society. If one could paraphrase Hegel, one could claim that each society has the journalists and the media it deserves. Media and journalism shape realities and nurture debate in thepublic sphere, strengthening democratic dialogue. However, the digitalization of the era, the technological advances in general, the turn towards the ‘new digital media” challenge journalism and journalists on multiple levels.

• •    Journalism is being transformed “in the ways it is produced, distributed and used”(Van der Haak, Parks & Castells, 2012). We observe both media convergence and a numerical diversity of media. “This nexus of ownership and market power spans different segments of the media and is qualitatively different from previous times” (Winseck, 2008).

More often than not, these companies hold interests beyond the journalism domain and are closely linked to political elites and power centers, thus challenging the independence of journalistic content, already limited by lesser source diversity. In the past decade journalism adapted to technological and societal changes and incorporated:

• •    new forms of presentation (multimedia story-telling, immersive journalism),

• •    new ways of involving the audience (participatory journalism),

• •    new ways of addressing issues and highlighting solutions (constructive journalism),

• •    and new forms of automation robot journalism and data journalism.

Kovach and Rosenstiel (2001) describe journalists as providing a public service: as watchdogs, gatekeepers, active collectors and disseminators of information; as striving for objectivity by remaining impartial, neutral, objective, fair and (thus) credible. Journalists must be autonomous, free and independent in their work. Moreover, journalists work with asense of immediacy, actuality and speed which is inherent in the concept of news; and last but not least journalists have a sense of ethics, validity and legitimacy.

In other words, journalism provides a set of values that relate to news, the truth and public service, despite of the huge diversity in the field. These values become meaningful in the news culture in a specific time and place. Journalists come to embody these values in their everyday practices at work, In the center of journalism practice lies the factor journalist, whose ethos and practices determine to a great degree of the quality of journalism presented. “The industry that has arisen around journalism’s everydayness does not define what it is–the idea(l)s, debates, and practices of journalists inhabiting these institutions do” notes Mark Deuze (2019).

Automated algorithmic content creation tools have been increasingly used in media across the journalistic spectrum. For the past few years organisations like the Washington Post or the GuardianadoptedAI to draft stories. News for the weather for example have been written by AI. The Associated Press (AP) uses this technology to produce content with Wordsmith, a software tool developed by Automated Insights, in what has been named as Algorithmic Journalism (Doer, 2015).Similarly, we have encountered terms such as robot journalism, automated journalism or machine written journalism (Anderson, 2012).

Three main ethical concerns regarding AI Journalism

• 1 .Human agency: human (moral) agency is partly at least delegated to the algorithms, (Dörr &Hollnbuchner, 2017) “allowing limited algorithmic intentions and autonomy”. Non journalistic actors are involved in the news

• 2 . Data reliability, objectivity and responsibility: The technical nature of coding, often requires the involvement of external providers. Next to the possibility of bias, questions regarding the reliability of the data, the methods used in data collection and processing as well as the legal issues regarding amongst others limitations and privacy should be considered. Transparency is a key concept in both the mining of data and the publishing the generated stories based on those data. Moreover, the fact that journalistic content spread on various media, reaches a vast audience of various demographics, makes the question who owns the data extremely important.

• 3 .ChatGPT and other similar programs have the capacity to generate and promote false narratives. Technology may be used to facilitate knowledge or disseminate misinformation. In an Open AIreport published in 2019, whose authors included GPT researchers warned of“three tiers of malicious actors:

production, thus changing the hierarchies within the media organizations and more importantly, changing the norms concerning accountability and bias. Who is accountable for the final product of the algorithmic journalism? Has the creator of the algorithm incorporated their own bias in it?Is the journalist in the position to interact and adapt the software?

• 1.    Low-skilled, limited resource actors who may be ideologically motivated or simply curious in their abilities. They may attempt to alter training data to bias a language model.

• 2.    Actors with moderate programming skills and resources who are able and willing to build a malicious product, such as tools for webspam.

• 3.    Advanced persistent threats (APTs): highly skilled and well-resourced groups, like state-sponsored actors, that have a long-term agenda” (Solaiman, et al., 2019).

According to the report the motives for all three actors could be the same, that is the pursuit of monetary gain, a particular political agenda, and/ or a desire to create chaos or confusion.

The principle of ChatGPT, to generate a response based on thousands of internet sources, makes it a valuable asset. The fact however that for the time being at least, copy generated by AI is imperceptible to human readers and anti-plagiarism software, makes it problematic. Authorship for any given content should be clearly stated, and even though that ChatCPT may be attributed as the author/source, that does not speak of the reliability of the content.

I have posed the question on ChatGPT regarding misinformation and ChatGPT. It generated a text in which it specifically stated: “it is important to note that as a language model, I do not have the ability to discern the accuracy of the information I process, I generate responses based on statistical patterns derived from the data I saw trained on, and my responses may reflect biases or inaccuracies that exist in that data [...] filtered and curated to minimize […] false or misleading information.”

It seems that algorithmic ethics are needed more than anything else. Nicholas Diakopoulos, Professor of Computational Journalism, writes that journalism should investigate the “societal power exerted through such algorithms”. “[…] various newsworthy angles on algorithms including discrimination and unfairness, errors and mistakes, social and legal norm violations and human misuse” (Diakopoulos, 2019).

Journalists ethical compass regarding the stories created by AI should be fully operational and focused. Beyond the expected questions of accuracy, right to the data journalists working with AI should be making the decision of which story should be lead or when an update or sudden development must be covered. Journalists should also be able to determine which subject matter is appropriate for automation and whether full disclosure of who (human or AI) is behind the content (Kent, 2015).

The ethical questions and challenges for journalism and journalists, arising from the development and utilization of algorithmic programs like ChatGPT, are real and require multifaceted responses. Journalists must develop competences in digital and computational literacy. Critical thinking and relationship building, fundamental to journalism, cannot be duped, or produced by algorithms. AI has a very specific role in journalism, to sort and integrate information, refine presentation, assist with distribution. The issue is:not whether data, computers and algorithms ca be used by journalists in the public interest but rather how, when where , why and by whom”(Howard, 2014). Latar (2015) points that “No robot journalist can become a guardian of democracy and human rights. It is therefore extremely important that human journalists should understand the dramatic developments in their professions and make sure these changes serve them in ways that will preserve and strengthen their very important social function.”

AI companies should implement ethical standards regarding their acquisition, interpretation and usage of data. Concerns about transparency, privacy bias and discrimination should be addressed promptly. A legal framework in which fundamentals rights and freedoms are safeguarded must be agreed upon, so that AI journalism can be “this approach to knowledge”, as Professor Shapiro (2020) declares.

4.    Conclusions

It has been clearly demonstrated that AI constitutes more a reality than a future goal since it is appeared in several areas of private and public life. Furthermore, European digital strategy encompasses the development of Artificial Intelligence in order to modernize digital environment and maximize the growth potential of the digital economy. What is challenging is to pursue the safe use of AI which would lead to the enhancement of digital culture and consciousness.

However, towards that road, there are many obstacles focusing on the lack of a legal binding legislation governing AI at european level. Meanwhile, serious ethical concerns arise since it is dubious whether an AI learning machine can substitute human thinking and behavior. Therefore, the crucial question is the following: how human-centric approach of AI can be achieved?

The intense and critical research through this paper verified the aforementioned issues. The central element of our study is the ChatGPT which constitutes the most recent model of the scientific and digital revolution. Its evolution and capacities are tremendous since its scope of application is extremely wide and can provide a large number of services. One month ago, a judge in Colombia has caused a stir by admitting he used the artificial intelligence tool ChatGPT when deciding whether an autistic child’s insurance should cover all of the costs of his medical treatment. He also used precedent from previous rulings to support his decision (Taylor, 2023). As it has been underlined at the beginning of the paper, according to the official statistics, in January 2023, ChatGPT users crossed 57 million, which furthered to over 100 million in February 2023. As it is explicitly declared, “ it had a recordbreaking adoption rate in the history of the tech industry. This tremendous growth is the result of massive word-of-mouth marketing! ” (Gohil, 2023).

Consequently, the constantly increasing rate of ChatGPT users makes its use inevitable and it is only a matter of time before its official and institutional adoption in public life sectors. The fact that a combination of algorithm and information collected and processed leads to a complete source of knowledge is really provocative as well as innovative.

However, many legal and ethical concerns arise, as it has been already demonstrated. Undoubtedly, the new AI learning machine tool must be regulated, according to its creator (Simons, 2023). This has been thoroughly analyzed through the present paper (Section I). In addition, the fact that its operation is based on the collection of information from an undefined number of persons also poses risks for the reliability of the research results. Therefore, this inherent weakness of the system, regarding the issue of reliability, which penetrates the whole function of the AI tool, can lead to an increase of fake news or disinformation. Thus, the need to control the content collected on the basis of which specific results are extracted, becomes even more imperative.

Apparently, it is our constant belief that living in the digital age presupposes trust in scientific evolution. Regarding AI it is of primary interest to achieve a powerful and trustful AI, adopting a human-centric approach. Pending the final adoption of the AI Act, in terms of the ChatGPT, any institution/organization which intends to apply the aforementioned learning machine tool shall previously design a specific data protection policy, with respect to fundamental rights and freedoms. In that vein, according to the GDPR provisions, the impact assessment is absolutely required.

Besides the legal requirements, regulating ChatGPT also needs to meet certain ethical standards in order to tackle against the ethical concerns already mentioned. Therefore, any authorization of use of this learning machine tool shall must be obtained after prior consultation of all parties involved. At this point it should be underlined that the establishment of a special consultative committee would be a reasonable and effective measure. That committee may be consisted of the following persons: a) the DPO of the institution/organization, b) the head of Human Resources Department and c) the Head of IT Unit. Undoubtedly, the participation of any other person expertized in the area of new technologies is considered necessary and valuable. The members of the committee should adopt an holistic and critical approach towards any case of potential use of ChatGPT, taking into account several principles such as those to necessity and proportionality and with respect to fundamental rights and freedoms. Finally, within the scope of its capacities, this special Consultative Committee would communicate with the competent national data protection authorities.

In terms of conclusion, it must be underlined that technology always precedes law. Consequently, any scientific evolution should be approached, in principle, positively. ChatGPT constitutes an extremely innovative AI learning machine tool providing amazingly unlimited services. Certainly, significant legal and ethical issues may be raised. However, as it has been thoroughly analyzed, there can be suggested several fruitful solutions based on strong legal background, following the recent legal and scientific evolutions. It is time to adopt an update digital culture, to incorporate in every European citizen a high level digital consciousness giving a special emphasis on the protection of human dignity. ChatGPT as any other AI model is generated by human thinking and as such it shall respect human’s rights and needs. We shall always remember that statement. Only then we would build a safe, trustful, powerful and gainful digital environment.

Kouroupis Konstantinos

Pravni fakultet Univerziteta Frederik, Kipar

Lambrou Evie

Fakultet za novinarstvo, komunikaciju i masovne medije Univerziteta Frederik, Kipar