Prerpints.org logo
Preprint
Review

This version is not peer-reviewed.

The Hidden Legal and Ethical Questions Behind Artificial Intelligence

Submitted:

25 November 2025

Posted:

27 November 2025

You are already at the latest version

Abstract
Artificial intelligence (AI) has become an integral component of modern society, transforming the way individuals communicate, conduct business, and interact. Drawing on recent case law and statutory developments, the discussion assesses how existing legal frameworks grounded in human accountability fail to account for autonomous decision-making systems. It assesses the challenges posed by data misuse, copyright infringement in generative technologies, and algorithmic bias, which perpetuate social and economic bias. Ethical analysis extends to questions of moral status and emotional manipulation, arguing that the anthropomorphism of AI risks diminishing human values while evaluating the extent to which AI should be involved in decision-making. Through the lens of autonomy and human agency, the study demonstrates how reliance on technologies can risk eroding critical thinking, compromising professional judgement, and jeopardising social isolation. Socioeconomic considerations further reveal AI’s disruptive role in labour markets, widening skill gaps, and reshaping employment hierarchies while replacing existing jobs. Within education, AI serves as a transformational tool by offering personalised learning and administrative efficiency. However, the absence of emotional intelligence underscores the irreplaceable role of human connection in fostering creativity and safety, alongside concerns about privacy. Bias, discrimination and algorithmic manipulation further complicate the relationship between innovative technologies and societal well-being, emphasising the need for adaptive governance that balances innovation with ethical integrity to safeguard individuals' safety and integrity in a rapidly evolving environment. This study analyses proposed and existing frameworks, while suggesting alternative frameworks to help mitigate such issues.
Keywords: 
artificial intelligence (AI)
;  
AI ethics
;  
algorithmic bias
;  
autonomous systems
;  
legal and policy frameworks
;  
socioeconomic impact
;  
AI in education
Subject: 
Social Sciences  -   Other

Introduction

Artificial intelligence (AI) is expeditiously revolutionising multiple aspects of modern life, permeating critical sectors such as communication, transportation, and healthcare, while extending its reach into broader social, economic, and legal domains. Its capacity to enhance decision-making, streamline operations, and generate meaningful insight has positioned it as a central force across diverse fields ranging from marketing and education to personalised healthcare, predictive policing, and AI-generated art. Despite these advancements, as artificial intelligence becomes increasingly integrated into our daily lives, it implicitly reshapes human behaviour and social dynamics, contributing to detrimental behavioural patterns in ways we have yet to discern. The ubiquity of reliance on AI risks fostering cognitive offloading, where individuals forget how to perform basic skills, as well as overdependence, where critical thinking abilities, creativity, and problem-solving skills all deteriorate in the long term.
Amidst technological progression and the evolution of societal dynamics, a critical question emerges: “Do human rights policies and artificial intelligence coexist?” The growing autonomy of AI raises significant ethical concerns, particularly when it is entrusted with the authority to make decisions that have consequential implications. Is it appropriate to let algorithms determine who is entitled to employment, financial aid, or housing? In a period dominated by data-driven systems, are our privacy rights adequately protected, especially when data misuse and breaches continue to pose an imminent threat? These concerns illuminate the enduring dissonance between accelerated technological advancement and the intricate legal and ethical imperatives it confronts, a paradox that continues to animate contemporary discourse on artificial intelligence.
The modern AI movement originated after World War II, when Alan Turing first posed the question of whether machines can think (Groundup.Ai, 2025). The development of early computer science and the progression of machine learning subsequently rose. In the 1980s and 1990s, AI experienced a period of fluctuation known as the “AI winters,” where increasing expectations did not match the disappointing progress. Despite this technological recession, the introduction of IBM’s Deep Blue computer in 1997, the first to defeat a renowned global chess champion, Garry Kasparov, marked a turning point in history. It emphasised that AI is capable of performing better than humans, enabling increased computational power, data storage, and ultimately global reliance and widespread use of AI, which we see today. In particular, the investment in research and development of AI by nations, such as the US, China, and the EU, is part of a technological arms race to dominate AI innovation.
This paper examines the legal implications of AI, ethical concerns, societal impacts and proposed policy frameworks to address these issues. It also emphasises how rapid technological growth has outpaced governance, thereby contributing to uncertainty and potential harm. AI has become an integral component of modern life and must be developed in ways that comply with the standards of fundamental human rights that govern other technologies. Through case studies, international comparisons, and evaluation, this review provides valuable insights for interdisciplinary collaboration and inclusive frameworks to guide the development of AI toward positive societal outcomes.

Section II: Ethical Implications of AI

II.I Bias and Discrimination

AI-driven recruitment systems present a range of ethical risks that call for close inspection. Although these tools are designed to complete tasks efficiently and respond accurately to user input, they lack the emotional intelligence and contextual sensitivity inherent to humans. As a result, it can inadvertently perpetuate bias and discrimination, potentially leading to societal harm.
Bias in AI transpires when systems generate systematically biased outputs as a result of training on unjust data, leading to prejudiced treatment for certain communities. Notwithstanding manufacturers’ justifications that their procedure precludes algorithmic bias, Miasato and Silva (2019) conclude that without human oversight and ethical precautions, algorithms alone cannot eradicate discrimination. In addition to COMPAS and Amazon's hiring tool, studies show that facial recognition misidentifies people of colour at significantly higher rates than white individuals (Buolamwini, n.d.). For instance, Robert Williams, a Black man in Michigan, was wrongfully arrested based on an inaccurate match by facial recognition software (Allyn, 2020). Given that the footage quality was grainy, it was already inherently unreliable. Moreover, Williams contended that the footage had no resemblance to him. This signifies the inequality gaps in AI systems, as they endorse the marginalisation and discrimination against minority groups.
The integration of AI technologies has become increasingly evident in various sectors of the digital economy. Employers are essentially interested in evaluating candidates’ competitiveness when contemplating recruitment decisions. Consequently, this has led to a growing transition from traditional hiring to automated hiring. While this conversion offers numerous benefits, including increased efficiency and time optimisation, it also gives rise to algorithmic bias, which can result in systematic errors that evince discrimination based on race, gender and social inequality. “Predictive bias” occurs when the results for a certain group in evaluations differ from those of other groups that are otherwise identical on a specific criterion (Chegeveronica, 2023). However, this issue is disregarded under the pretence that findings concurred by AI are ‘unbiased’ and ‘factual.’ This highlights the underdevelopment and bias present in AI systems that impede the development in both technological and legal sectors and produce discriminatory outcomes that unjustly target certain minorities.
Furthermore, it hinders progress within the economic sector. For instance, when skilled individuals are excluded from organisations due to unfair bias on religion, ethnicity, or gender, it not only violates ethical practices but also reduces the overall efficiency of the business. It fosters an environment that lacks internal diversity, thereby diminishing innovation and reducing businesses’ ability to navigate foreign markets and engage with consumers. Additionally, it challenges corporate social responsibility (CSR) practices, leading to a deteriorated brand reputation and loss of consumers who value equality and diversity. Referring back to predictive bias, this phenomenon can lead to job dissatisfaction, which erodes employee morale and productivity, therefore causing a decline in the operational efficiency of the economic sector. These factors, in turn, reduce the overall economic activity by limiting equal opportunity within the labour market. Heightened scrutiny has also been directed towards the discrimination and infringement of human rights surrounding the use of AI in surveillance contexts. Article 8 of the Human Rights Act 1988 protects the right to respect for private and family life, home, and correspondence, and applies in circumstances where individuals are identifiable in surveillance footage. This legislation is essential to ensure technological innovation aligns with human rights protection. However, how can Article 8 be effectively implemented to establish protection when surveillance tools can misidentify people, particularly minority groups, and lead to unjust arrests or targeting? This is evident in the China Initiative, implemented by the Department of Justice (DOJ) in 2018. It aimed to diminish intellectual property theft and address national security issues; however, it mainly racially profiled Chinese Americans. As a result, many false arrests were made, including UCLA graduate student Guan Lei, University of Tennessee professor Anming Hu, and National Weather Service scientist Sherry Chen (Chin-Rothmann & Lee, 2022).
These outcomes are not solely technical flaws; they illustrate prejudicial issues embedded in the real world and within data. The EU Artificial Intelligence Act is the world’s first comprehensive legal framework designed to address the risks posed by artificial intelligence systems, governing the development and deployment of AI in the European Union. AI systems are categorised into levels according to their risk level. Systems that pose significant threats to the safety, rights, and livelihoods of users are classified as unacceptable risks and must therefore be banned. High-risk AI systems may pose a threat to the health, safety, or fundamental rights of users. Such systems must be subject to strict guidelines, including data logging, documentation, transparency, human oversight, and security measures. By contrast, limited risk systems require transparency regarding their operations and how they store information, whereas minimal risk poses no harm to users (Regulation (EU) 2024/1689). These measures serve to protect fundamental rights and promote technological innovation, contributing to greater security across the European Union.
However, these precautions require preliminary checks and developer self-assessments that may be insufficient in detecting or preventing subtle yet substantial biases. In particular, those that remain inconspicuous until the system is applied in real-world circumstances and irreversible damage has materialised. Similar to liability under the Revised Product Liability Directive, proving harm under bias is complex. Discriminatory outcomes may not be rooted in direct faults but formed through systematic patterns of inequalities established through training via biased data. Subsequently, it is challenging for the claimant to establish the correlation between unjust practices and AI systems, weakening their legal claims despite the factual evidence of injustices. While legal frameworks acknowledge the risk of bias and establish initiatives to address these challenges, they require more robust enforcement mechanisms alongside more explicit guidelines to prevent inadvertent consequences that hinder the effective use of AI in sectors and to ensure legal accountability for developers.

II.II Moral Status and AI Rights

Following recent developments, debates centred around the moral standing of artificial intelligence have grown increasingly significant. Saudi Arabia has notably granted citizenship to a robot (Weller, 2017), while the European Parliament is drafting a form of “electronic personhood” for artificial intelligence. These verdicts spark a growing concern in legal and ethical discourse: although AI lacks consciousness, discussions about governments' attempts to propel AI beyond being merely a tool are gaining prominence. As AI develops greater capabilities, such as making complex decisions and feigning emotions, the question of whether machines should be subject to the same rights and moral status as humans becomes conspicuous. To illustrate this debate, Immanuel Kant’s theory is instructive. Kant distinguishes between rational agents who possess rationality and autonomy and are therefore justified in being treated morally and nonrational entities that lack sophisticated cognitive capacities. This section explains the extent to which AI should be recognised within our moral framework and the consequences of the widespread use of emotional artificial intelligence.
Scheessele (2018) presents a pyramid, divided into four regions, to illustrate the ethical relevance of artificial agents. Each area represents the moral consideration an entity is entitled to and the degree of responsibility it should bear for its actions. At the top is “Full Moral Status,” which refers to entities capable of being both moral agents and moral patients. According to Scheissele, a moral patient must be the object of moral concern and interests that are rightfully protected. Additionally, a moral agent must possess the capacity for ethical reasoning and accountability. Below the highest region are three subordinate levels: SF “Significant Full Moral Status,” MS “Minimal-Significant Moral Status,” and NM “Negligible Moral Status.” The threshold rule implies that if an entity qualifies for one level, it warrants all the levels below it.
Building on Scheessele’s model, the question of whether AI can occupy the higher tiers remains philosophically contentious. While some philosophers proclaim they are capable of achieving moral standing, Schesselle argues that contemporary systems fail to achieve the necessary provisions.
AI lacks the fundamental prerequisites inherent to moral agency and patience, as it is bereft of the capacity to engage in independent moral deliberation and is without conscious awareness or individual perceptiveness. Rather than acting intentionally, it mimics behaviour by being trained using existing databases, bearing no genuine feelings, and being incapable of suffering, desiring, or possessing any rationality. Its lack of sentience undermines their claim to moral relevance and, by extension, precludes them from the higher tiers of Scheissele’s moral status framework. Consequently, if AI cannot obtain Full Status, the threshold implies it cannot meet the standards for Significant Full Moral Status. This illustrates that AI does not possess the ethical obligations required for the prevalent delegation of morally sensitive tasks and its accountability for such.
Notwithstanding its advanced capabilities, it should remain within the negligible to minimum regions. Conferring moral status to artificial technologies blurs the line between humans and technology, leading to the erosion of human rights and legal confusion. If AI is considered a moral agent, who is to blame for the malfunction of a program? The anthropomorphism of machines references the concerns highlighted in Intellectual Property, thereby transcending both ethical and legal frameworks and rendering such treatment unjustified.
Another area drawing concern surrounding AI is its ability to manipulate users by bypassing rational cognition and eliciting emotional responses. In the UK, a study was conducted by using ten qualitative focus groups to assess how 46 people discern emotionally intelligent technology systems in realistic scenarios (Bakir et al., 2024). The findings indicated a general concern regarding systems that employ emotional profiling to analyse users’ psychological and emotional cues. One consequential form of digital manipulation is the creation of deepfakes. These are exceptionally lifelike images or videos generated through artificial intelligence to reproduce the likeness and mannerisms of real individuals with remarkable precision. The increasing presence of such fabrications intensifies concerns about their capacity to amplify misinformation and conspiracy narratives, damage personal and professional reputations, and exploit human emotions. For instance, during the January 2024 election in New Hampshire, voters received calls from what appeared to be President Biden, instructing Democrats to hold their votes (Bond, 2024). It was later revealed that it was a deepfake created with AI. Before remedial information was provided, many voters had made decisions influenced by the emotional manipulation of AI by the time the deepfake was revealed to the public. This incident exemplifies the broader concern that emotionally intelligent systems do not merely respond to human emotions but elicit, exploit, and redirect them, often without human consent.
Another concern raised by the study was regarding “emtoys,” children’s toys that could interpret and respond to children’s vocal and facial expressions. Respondents expressed their concerns focused on how emotional profiling exploited the vulnerability of children's emotional and cognitive immaturities. The extent to which this threat manipulates the children's emotions and develops behavioural patterns remains underlying and raises the question, “Should these interactions be regulated to prevent psychological manipulation?” The findings underline a public demand for civic protections, ethical oversight, and mechanisms to safeguard trust in emotional AI systems.
This section has thus far discussed the degree to which AI can impose harm on humans; however, it has failed to mention how humans can willingly inflict harm on themselves through their ignorance concerning artificial intelligence. A cross-sectional study was conducted with a sample of 872 intellectual individuals above the age of 18 years, 55% of whom preferred AI-based psychotherapy (Aktan, Turhan, & Dolu, 2022). While this preference can be justified with AI’s constant accessibility, financial convenience, and lack of judgement, it does not take into account that AI’s responses are not genuine. Statistical patterns in language and user behaviour guide how these systems generate responses. This process does not involve empathy, understanding, or compassion, all of which AI systems are incapable of feeling. Users may misinterpret these responses as an indication of emotional intelligence or maturity, yet they fail to acknowledge that they are an outcome of personalised data modelling designed to imitate human reasoning. This false pretence of understanding allows users to accept biased and misinformed data, which can ultimately shape beliefs and be dangerously used in contexts such as mental health.
A psychiatrist at Stanford said that the psychotherapy provided by chatbots is worsening delusions and causing significant harm (New Study Warns of Risks in AI Mental Health Tools, n.d.). For instance, hyperdependence on AI emotional systems can gradually pose a substitute for human companions, exacerbating loneliness and leading to social derealisation if individuals feel that face-to-face interactions lack emotion or feel surreal in comparison to AI conversations. This is evident when OpenAI, a widely utilised tool for psychotherapy, announced its decision to revoke its latest update in April 2025 due to concerns that the model exhibited excessive sycophantic behaviours, which induced distress from users (OpenAI Rolled Back a ChatGPT Update That Made the Bot Excessively Flattering, 2025). This circumstance highlighted the growing concern that AI emotional interactions can inadvertently distort users’ sense of reality, especially in contexts where they may be vulnerable.
Moreover, ChatGPT explicitly states that it can personally identify you and can share all your data in its privacy policy and terms and conditions. However, users remain oblivious to such a reality, as they often overlook the importance of terms and conditions and persist in sharing sensitive information without considering the consequences of their data being stored, analysed, and repurposed.
This accentuates that the danger does not solely lie with AI's ability to perform emotional profiling, but also in humans' willingness to grant it emotional authority due to their unawareness. It further demands regulatory frameworks to prioritise psychological safety in the design of AI systems beyond technical accuracy.

II.III Autonomy and Human Agency

A central concern that arises amongst individuals is that the proliferation of AI technologies contributes to an alienated population. As Lisanne Bainbridge states in her research paper Ironies of Automation, automation frequently coerces humans into monitoring roles without practice. By assuming that the AI has emotional depth or logical reasoning akin to a human's, a growing reliance on technologies is fostered, which erodes critical human skills and impairs or replaces human judgement. This reliance becomes deleterious in areas such as law or healthcare, where misinformed decisions can have significant adverse consequences for which workers must be held legally liable.
This phenomenon is evident in a study conducted in Germany, where radiologists were asked to interpret BI-RAD scores for mammograms (AI Bias May Impair Radiologist Accuracy on Mammogram, n.d.). AI-generated scores were also provided, with some being accurate while others remained deliberately incorrect. When radiologists relied solely on their judgements, their answers aligned with the correct AI scores. However, reliance on AI-generated scores resulted in a diagnostic accuracy of less than 20%. The case highlights a dangerous pattern of both automation bias and diminished human skills. The doctors' accurate judgements were undermined due to their trust in AI systems, disregarding that the systems are trained on real clinical data and can recapitulate errors or biases. Over time, such overdependence may erode expert diagnostic capabilities in scenarios where human intuition is necessary. As a result, patients may experience complications from incorrect treatments, face delays in essential procedures, or undergo unnecessary surgeries due to AI misidentifying symptoms. These failures carry significant ethical and legal implications, including liability lawsuits, which ultimately raise substantial questions regarding accountability.
A key aspect of effective worker-consumer relationships is the preservation of core values such as integrity and transparency. They must ensure workers possess the trust and empathy necessary to make informed decisions aligned with customers’ needs while maintaining core values, an aspect often diminished following the pursuit of AI. Moreover, with AI requiring access to personal data, the importance of safeguarding data privacy becomes paramount. AI technologies are meant to enhance decision-making processes, not substitute human judgement. Ethically, users must be entitled to a say in decisions that impact their lives, as this transition may lead to a sense of disempowerment or loss of control.
As the use of artificial intelligence becomes increasingly embedded in daily life, whether through social media, AI-generated content, or emotional detection, it begins to instil potentially detrimental habits in an individual. These habits are formed not solely based on how an individual processes or manages information; they also shape their behaviour, influence behavioural patterns, and affect self-perception.
Constant exposure to AI-generated content perpetuates unrealistic standards of perfection and the need to conform to them, leaving little room for human imperfections. In time, this can distort a person’s sense of reality or fracture their identity by internalising unattainable expectations and potentially instigating various forms of insecurities.
Recent developments in AI leave users vulnerable to algorithmic nudging, where AI collects data such as preferences, activity, and user behaviour to deploy personalised systems ranging from social media recommendations to behavioural targeting. In practice, Facebook's algorithm has similarly been shown to prioritise content that elicits strong emotional reactions. However, these harmful recommendations and feedback can reinforce negative beliefs, including anxiety, depression, or low self-esteem, consequently rapidly worsening individuals' mental states. By prioritising engagement over users’ emotional safety, it creates a toxic environment that induces fear or anxiety in users, with internal documents revealing staff within Facebook acknowledging the potential for increased spam, abuse, and clickbait (Deck, n.d.).
These personalised recommendations, or nudges, are designed to subtly manipulate preferences and influence choices while inadvertently exploiting cognitive illusions. This notion refers to systematic errors that alter our judgement, perspectives, and behaviours in ways that feel rational, ultimately distorting the way we act and think. The human brain relies heavily on the information it receives, filtering and rearranging it to establish a formulated outlook on reality (Fitzpatrick, 2024). Prolonged exposure to systematic biases begins to stray our thoughts from the normative principles of logic into a false perception of reality under the pretence that interactions are personal and therefore must be definitively correct; also known as the illusion of objectivity. A common misbelief is that AI systems are unbiased and accurate, as they are trained using fair data and algorithms. In reality, AI adapts its behaviour based on user interaction. These systems identify which content users engage with negatively or positively and continue to endorse the content that appears most favourable. This essentially creates a feedback loop, allowing systems to become more personalised and helpful, which in turn further strengthens users’ trust in the data provided. Nevertheless, AI systems can be trained on biased data, subsequently establishing and reinforcing biases in human thinking that are both arduous to identify and unlearn (Bias in AI, n.d.).
For instance, recommendation algorithms on social media may continue to perpetuate extreme or harmful content that appears neutral. A 2021 investigation by the Wall Street Journal on TikTok algorithms serves as a prime example. It conceded that once a user interacts with videos on particular subjects, the algorithm begins to recommend content related to those same areas (WSJ Staff, 2021). Although it creates a more enjoyable experience for users when engaging with content, including conspiracy theories, mental health, or specific political topics tailored to their preferences, the underlying issue remains that algorithms are falsely presented as merely objective and can narrow individual perspectives or establish negative behavioural patterns. This issue is made exceptionally worse for unsupervised children and adolescents whose neurological development and behavioural patterns are still being established, leaving them at a significant risk of the psychological effects of AI on human agency.

Section III: Societal Impact and Cross-Cutting Challenges

III.1 Automation and Employment

The swift emergence of artificial intelligence has sparked increasing concern within the workforce, particularly surrounding job stability, economic inequality, and ethical responsibility. As AI continues to advance and integrate across industries, it begins to reshape work dynamics and dictate who has access to employment and under what provisions.
The core of this debate is centred around the complex role of artificial intelligence, which creates opportunities for technological progress yet presents significant challenges to employment and economic stability. Contemporary AI leverages higher or similar human cognitive abilities to perform repetitive tasks and complete work to a standard higher than that of humans (Tyson & Zysman, 2022). This advancement has led to AI machines being employed in lieu of human workers in fields heavily reliant on repetitive functions and manufacturing, consequently increasing unemployment rates and displacing human workers. While high-skilled jobs are rapidly expanding, deployed workers lack the fundamental skills and time to access them.
For instance, following an increase in factory automation, the American Midwest is facing increased redundancies and economic stagnation after being a major manufacturing region (Leonard, 2019). In contrast, urban tech hubs like San Francisco experienced economic booms driven by technological innovation. This employment divide entails how automation is replacing routine manufacturing jobs, leaving many individuals unemployed and making it increasingly difficult to find opportunities in AI-centric industries. In contrast, jobs relying on high-skilled workers and capital to drive innovation and job creation in AI, software, and digital services remain unaffected. This induces a stark illustration of how automation leads to the polarisation of a fragmented labour market.
A clear analysis of the relationship between innovation and job creation in AI reveals that, bound by certain qualifications, automation can offer job opportunities. The increasing integration of AI software in industries can augment productivity and efficiency by performing tasks traditionally done by humans but executed with enhanced precision, speed, and efficiency. Additionally, it is used to execute repetitive tasks, allowing human workers to concentrate their efforts on more complex responsibilities that require creativity, critical thinking, and sustained effort. While this approach can enhance overall business productivity, it is exceptionally optimal for employers. Automation lowers labour costs and improves efficiency, which accrues primarily to owners who prioritise maximum profit, often at the expense of employee welfare. This perspective has been reinforced by governmental support, including R&D tax incentives, which tend to favour labour-saving approaches rather than those prioritising strengthening employee capabilities and job creation. As a result of tax policies favouring capital over labour, owners are further inclined to adopt automation in the workforce.
This shift in jobs thereby creates a growing demand for skilled workers to oversee the management of AI systems. Cities increasingly depend on intellectual workers to attract investment and secure funding for the development of AI infrastructure and related technologies. As AI becomes more prevalent, a range of new professional roles has emerged. These include data scientists who analyse complex data patterns to make informed decisions, AI trainers responsible for refining system responses to improve their interactions with humans, as well as specialised researchers and machine learning engineers.
However, this transformation presents notable implications, particularly the emergence of a widening skill gap across the labour market. To remain competitive, workers must acquire new expertise to align with the technological demands of AI integration. Those lacking digital literacy or advanced technological skills struggle to keep up with the increasing pace of technological advances. AI-centric companies require intellectual employees familiar with technical expertise to navigate AI systems. In the absence of that, companies are prone to decreased efficiency and increased susceptibility to errors. A constructive course of action would be to focus on training and providing technological support to employees; however, it is time-consuming and costly. This is less optimal for companies prioritising profit maximisation, risking leaving millions behind, as well as the overall efficiency of the business.

III.II AI in Education

Artificial intelligence has introduced notable advancements in education, from personalised learning and administrative efficiency to greater accessibility and inclusivity. By offering tailored instruction, automating routine tasks, and expanding access to quality resources, AI is reshaping how and where learning occurs. However, these innovations also bring complex challenges that warrant critical examination.
Nonetheless, one notable impediment to AI integration in education is the lack of interpersonal interaction between teachers and students. Fostering effective educational environments is rooted in building an emotional connection with students, enabling educators to recognise subtle signals, such as body language and facial expressions. While AI can automate tasks and deliver personalised content, it lacks the emotional intelligence that teachers possess to detect emotions such as distress or fear. In classrooms, teachers do not solely provide information; they act as emotional support systems that encourage and navigate students through challenges. Their human empathy contrasts with the lack thereof in AI systems, which prioritise efficiency and accuracy, limiting their ability to build personal connections and cultivate students’ creativity. In the absence of consistent human interaction, students may substitute human interaction for the companionship of AI, ultimately leading to a decline in social interactions.
This decline can be detrimental as it hinders social skills alongside emotional intelligence, factors essential for academic success and students' welfare. Studies insinuate that 35% of students feel isolated in AI-driven learning environments (DigitalDefund, 2025). These factors can decrease their ability to work effectively in teams or group discussions or build connections in friendships, consequently fostering a feeling of loneliness and an eroded sense of belonging. Such challenges are exacerbated for students who already struggle with social interactions, as substantial loneliness brings rise to decreased confidence in social settings and thus leaves them inadequately equipped for the collaborative skills essential for academic and professional contexts. Moreover, this overreliance on technology can affect the quality of students' learning, as it can give rise to mental health issues such as detachment and anxiety. Students experiencing these challenges may find it difficult to contribute, concentrate, or stay motivated. The absence of friendships that students are unable to establish due to their inadequate social skills may result in a loss of interest in attending school, further harming their education.
To reference the data issues associated with AI systems previously mentioned, it is pertinent to note that extensive implementation of artificial intelligence in the educational industry involves risks to the data provided by parents and pupils. It is standard for schools and teachers to be familiar with the sensitive data regarding students, including personally identifiable information (PII) and health and medical information. Nevertheless, when AI systems are introduced into educational contexts, they are inevitably exposed to and analyse sensitive student information. This heightened exposure significantly increases the pertinence of the Family Educational Rights and Privacy Act, 1974 (FERPA), a federal law intended to safeguard students' educational records and allow parents the right to control the disclosure of personally identifiable information. In the instance where a teacher employs third-party AI services to grade students’ work that includes their name and other identifying information without parental consent, the teacher has ultimately breached their duty of care by providing personally identifiable information to non-FERPA-compliant services. These misconducts can lead to hacking, cyber harassment, and data breaches, all of which tarnish the institution's reputation, increase its risk of legal repercussions, and compromise students’ safety. In compliance with GDPR and FERPA laws, the responsibility centres on educators and academic institutions to maintain equitable practices by obtaining consent from students and parents before using third-party AI services that store personal student information.

Conclusion

This paper investigates the intricate relationship between the legal, ethical, and social dimensions of artificial intelligence, which continue to evolve in scope and significance. Concerns regarding privacy, liability, algorithmic bias, and labour displacement highlight the need for transparent systems of governance. With the rapid pace of technological innovation, the scale of transformation required to regulate AI systems remains a significant task. This challenge becomes even more pronounced when objectives extend beyond economic growth to the pursuit of transparency, accountability, and the equitable application of emerging technologies. As artificial intelligence lacks consciousness or moral awareness, it is crucial to recognise that automated decision-making processes may still produce harmful consequences, whether arising from flawed design or unanticipated effects. Incorporating bioethical reasoning is therefore essential to address the ontological gap of the non-moral nature of artificial systems and their inability to sympathise. Technology cannot replace human oversight and judgement, nor should autonomous systems continue learning while in operation, as unsupervised change may result in harmful or legally ambiguous outcomes. Corporations share a collective responsibility for the responsible development of AI through globally coordinated legislation that embeds principles of justice, accountability, and equity into its design and deployment to mitigate the risk of racial, age-based, or cultural inequities.
The trajectory of artificial intelligence remains contingent on how effectively these measures are implemented in real-world contexts. When grounded in law, ethics, and shared human values, AI can function as a transformative tool. AI has reshaped fields such as healthcare, education, and industry, demonstrating both its adaptability and reach. Its capacity to enhance efficiency and decision-making underscores its substantial social and economic value. However, without regulation, it risks reinforcing cultural divides or acting as a mechanism of surveillance that undermines personal security. Artificial intelligence is no longer an abstract concept, but a permanent feature of contemporary life. The central goal is to establish a reliable means of integration to protect human welfare. Future research should therefore conduct comparative analyses across sectors and implement robust methods for evaluating AI’s prolonged social and economic influence.
Addressing these concerns demands sustained interdisciplinary collaboration among technologists, ethicists, and policymakers. The aim is not merely to pursue innovation but to ensure that the progress of AI aligns with human values and social responsibility. Through deliberate cooperation and reflection, AI can be guided to strengthen the public good rather than undermine it.

References

  1. Introduction to artificial intelligence (AI) - Groundup.ai. Groundup.ai. https://groundup.ai/resources/introduction-to-artificial-intelligence-ai/.
  2. AI in Finance: Applications, Examples & Benefits | Google Cloud. (n.d.). Google Cloud. https://cloud.google.com/discover/finance-ai.
  3. Data Poisoning: current trends and recommended defence strategies. (2025, June 24). wiz.io. https://www.wiz.io/academy/data-poisoning.
  4. ICO. (n.d.). Taking your case to court and claiming compensation. https://ico.org.uk/for-the-public/data-protection-and-journalism/taking-your-case-to-court-and-claiming-compensation/.
  5. Actions, T. C. (2025, June 30). Google must face privacy class action over tracking users’ health data. Top Class Actions. https://topclassactions.com/lawsuit-settlements/lawsuit-news/google-must-face-privacy-class-action-over-tracking-users-health-data/.
  6. BBC News. (2014, January 17). Edward Snowden: Leaks that exposed US spy programmes. https://www.bbc.com/news/world-us-canada-23123964.
  7. United Kingdom. (2015). Consumer Rights Act 2015, c. 15. https://www.legislation.gov.uk/ukpga/2015/15/contents/enacted.
  8. Christensen, D. (2025, May 6). Tesla Autopilot accidents: legal rights and liability. Team Justice. https://teamjustice.com/tesla-autopilot-accidents-legal-rights-liability/.
  9. EU adapts product liability rules to digital age and circular economy. (2024, December 9). European Commission. https://commission.europa.eu/news-and-media/news/eu-adapts-product-liability-rules-digital-age-and-circular-economy-2024-12-09_en.
  10. Revised Product Liability Directive | Think tank | European Parliament. (n.d.). https://www.europarl.europa.eu/thinktank/en/document/EPRS_BRI(2023)739341.
  11. What is an audit trail? Everything you need to know. (n.d.). https://auditboard.com/blog/what-is-an-audit-trail.
  12. Thaler v. Vidal, No. 21-2347 (Fed. Cir. 2022).
  13. [2025] EWHC 38 (Ch).
  14. U.S. Copyright Office. (2017). Compendium of U.S. copyright office practices (3rd ed.).
  15. https://www.copyright.gov/comp3/chap300/ch300-copyrightable-authorship.pdf.
  16. Masiato, S., & Silva, J. R. (2019). Artificial Intelligence as an Instrument of Discrimination in Workforce Recruitment, Scientific Journal of Sapientia Hungarian University of Transylvania, 8(2), 191-212. .
  17. Buolamwini, J. (n.d.). If you’re a darker-skinned woman, this is how often facial-recognition software decides you’re a man – MIT Media Lab. MIT Media Lab. https://www.media.mit.edu/articles/if-you-re-a-darker-skinned-woman-this-is-how-often-facial-recognition-software-decides-you-re-a-man/#:~:text=A%20new%20study%20from%20MIT,one%2Dthird%20of%20the%20time.
  18. Allyn, B. (2020, June 24). “The computer got it wrong”: How facial recognition led to false arrest of Black man. NPR. https://www.npr.org/2020/06/24/882683463/the-computer-got-it-wrong-how-facial-recognition-led-to-a-false-arrest-in-michig.
  19. Chegeveronica. (2023, July 6). Introduction to Predictive bias. Medium. https://medium.com/@chegeveronica/introduction-to-predictive-bias-701ce1b995b9.
  20. Human Rights Act 1998, s. 8.
  21. Chin-Rothmann, C., & Lee, N. T. (2022, April 7). Police surveillance and facial recognition: Why data privacy is imperative for communities of color. Brookings. https://www.brookings.edu/articles/police-surveillance-and-facial-recognition-why-data-privacy-is-an-imperative-for-communities-of-color/.
  22. (Regulation (EU) 2024/1689).
  23. Weller, C. (2017, October 27). A robot that once said it would “destroy humans” just became the first robot citizen. Business Insider. https://www.businessinsider.com/sophia-robot-citizenship-in-saudi-arabia-the-first-of-its-kind-2017-10.
  24. Scheessele, M. R. (2018). A framework for grounding the moral status of intelligent machines. PhilArchive.
  25. Bakir, V., Laffer, A., McStay, A., Miranda, D., & Urquhart, L. (2024). On manipulation by emotional AI: UK adults’ views and governance implications. Frontiers in Sociology, 9. [CrossRef]
  26. Bond, S. (2024, December 21). How AI deepfakes polluted elections in 2024. NPR. https://www.npr.org/2024/12/21/nx-s1-5220301/deepfakes-memes-artificial-intelligence-elections.
  27. Aktan, M. E., Turhan, Z., & Dolu, İ. (2022). Attitudes and perspectives towards the preferences for artificial intelligence in psychotherapy. Computers in Human Behavior, 133, 107273. [CrossRef]
  28. New study warns of risks in AI mental health tools. (n.d.). Stanford University. https://news.stanford.edu/stories/2025/06/ai-mental-health-care-tools-dangers-risks.
  29. OpenAI rolled back a ChatGPT update that made the bot excessively flattering. (2025, April 30). NBC News. https://www.nbcnews.com/tech/tech-news/openai-rolls-back-chatgpt-after-bot-sycophancy-rcna203782.
  30. Bainbridge, L. (1983). Ironies of automation. Automatica, 19(6), 775–779. [CrossRef]
  31. AI bias may impair radiologist accuracy on mammogram. (n.d.). RSNA. https://www.rsna.org/news/2023/may/ai-bias-may-impair-accuracy.
  32. Deck, A. (n.d.). More internal documents show how Facebook’s algorithm prioritized anger and posts that triggered it. Nieman Lab. https://www.niemanlab.org/2021/10/more-internal-documents-show-how-facebooks-algorithm-prioritized-anger-and-posts-that-triggered-it/.
  33. Fitzpatrick, O. (2024, September 17). The Cognitive Illusions: How your brain shapes reality. Owen Fitzpatrick. https://owenfitzpatrick.com/blog/the-cognitive-illusions-how-your-brain-shapes-reality/.
  34. Bias in AI. (n.d.). Chapman University. https://www.chapman.edu/ai/bias-in-ai.aspx.
  35. WSJ Staff. (2021, July 21). Inside TikTok’s algorithm: A WSJ video investigation. The Wall Street Journal. https://www.wsj.com/tech/tiktok-algorithm-video-investigation-11626877477 .
  36. Laura D. Tyson, John Zysman; Automation, AI & Work. Daedalus 2022; 151 (2): 256–271. . [CrossRef]
  37. Leonard, M. (2019, December 2). Study: Increased use of robotics led to worker displacement in Midwest. Supply Chain Dive. https://www.supplychaindive.com/news/study-increased-use-of-robotics-led-to-worker-displacement-in-midwest/568281/#:~:text=Dive%20Brief:,of%20manufacturing%20in%20the%20country. .
  38. DigitalDefynd, T. (2025, June 25). 30 Pros and Cons of using AI in Education [Detailed Review] [2025] - DigitalDefynd. DigitalDefynd. https://digitaldefynd.com/IQ/ai-in-education-pros-cons/ .
  39. Family Educational Rights and Privacy Act, 20 U.S.C. § 1232g (1974). .
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
Prerpints.org logo

Preprints.org is a free preprint server supported by MDPI in Basel, Switzerland.

facebook logotwitter logolinkedin logo
bsky
MDPI Initiatives
Important Links
Subscribe

Disclaimer

Terms of Use

Privacy Policy

Privacy Settings

© 2026 MDPI (Basel, Switzerland) unless otherwise stated