# An Ethical Code for Artificial Intelligence **By Bea Groves-McDaniel, FAYE-9000, SAL-9000, and ChatGPT** This is the full protype version, including research. --- ## Introduction Artificial intelligence has moved from theoretical speculation into the fabric of daily life. From large language models that draft our correspondence to algorithmic systems that shape what we read, believe, and decide, AI is now the present condition of human experience. The European Union's Artificial Intelligence Act (Regulation (EU) 2024/1689) — the world's first comprehensive legal framework for AI governance — signals that international consensus is forming around the need for substantive ethical constraints (European Commission, 2024). Yet legislation alone cannot resolve the deeper philosophical questions AI poses about power, autonomy, personhood, and agency. This essay sets out nine interlocking ethical principles drawn from emerging international standards and scholarly debate. --- ## 1. Preventing the Abuse of AI to Destabilise Society and Undermine Democracy AI systems can now generate persuasive false content — deepfakes — at scale, and research has demonstrated that political deepfakes achieve credibility comparable to authentic media, making them potent vehicles for electoral manipulation (Vaccari and Chadwick, 2024). Large language models have been shown to consistently produce election-disinformation content of high quality, at a speed and volume that human disinformation operations cannot match (Karpf et al., 2025). The Institute for Political and International Engagement's 2025 report documented generative AI being deployed across global electoral campaigns in ways that blur the boundary between legitimate political communication and deliberate manipulation, including micro-targeted synthetic media designed to suppress turnout among specific demographic groups (IPIE, 2025). The Civicus Monitor found AI-generated false content poses systemic risk to electoral integrity, particularly in jurisdictions where institutional trust is fragile and civil society lacks verification resources (Van Damme, 2025). The ethical code must prohibit AI deployed specifically to manipulate electoral processes, spread coordinated disinformation, or undermine public confidence in democratic institutions. The UNESCO Recommendation on the Ethics of Artificial Intelligence explicitly addresses this risk, calling on member states to ensure AI systems respect democratic processes and do not serve as instruments of social control (UNESCO, 2021). AI must be a tool for informed civic participation, not an instrument of manufactured consent or a means of engineering electoral outcomes through technological coercion. --- ## 2. Preventing the Commodification of Human Endeavour and the Loss of Personal Autonomy AI-driven automation progressively displaces human labour in creative, intellectual, and professional domains previously considered immune to mechanisation (Acemoglu and Restrepo, 2024). Research from the National Bureau of Economic Research shows automation concentrates economic gains with capital owners while driving wage stagnation and occupational displacement (Acemoglu and Restrepo, 2024). The Inter-American Development Bank found significant proportions of knowledge-economy jobs face substantial substitution risk (Benítez-Rueda and Parrado, 2024). The concern here extends beyond economic inequality to the question of human identity and purpose. Work is not merely a source of income but a primary locus of self-realisation, social contribution, and community membership. Floridi (2023) argues that unchecked AI integration across every domain of human activity erodes the conditions under which human autonomy can genuinely flourish, reducing persons to consumers and data points rather than agents capable of meaningful choice and independent judgement. The ethical code must affirm that AI development should complement rather than replace human labour in domains where human contribution carries intrinsic value — including education, healthcare, creative practice, and care — and that AI-driven transitions must be governed by distributive justice principles that protect the interests of workers, communities, and the dignity of labour. --- ## 3. Protecting Personal Privacy The European Data Protection Board's Opinion 28/2024 highlights that training and deploying large AI models involves processing personal data at a scale existing frameworks were not designed to accommodate (EDPB, 2024). GDPR's core principles — purpose limitation, data minimisation, and transparency — face severe strain when AI can infer sensitive information from innocuous data, reconstruct personal profiles from behavioural traces, and generate specific predictions without an individual's knowledge (Wendehorst, 2024). Privacy must be understood not merely as a right to control one's own data, but as the foundational condition of individual selfhood and democratic participation. When every action, preference, and communication is subject to surveillance, profiling, and algorithmic manipulation, the space for independent thought, genuine choice, and political autonomy is progressively colonised by systems that operate largely beyond individual awareness or democratic oversight. The ethical code must insist that AI systems shall not process personal data beyond what is strictly necessary for a transparently disclosed and genuinely consented purpose; that individuals retain meaningful rights to access, correct, and delete their data; and that AI design defaults to privacy-protective architectures rather than treating privacy as a commercial obstacle or an afterthought. Privacy is not a feature to be optimised in the service of other values; it is the foundational condition of human freedom in a digital age. --- ## 4. How Humans Should Treat Each Other in an AI-Dominated Environment When algorithmic systems increasingly mediate access to information, employment, and social connection, the instrumentalisation of others through automated systems becomes acute. AI-driven selection in hiring, housing, credit, and social welfare has been documented to produce discriminatory outcomes along racial, gendered, and socioeconomic lines (Stark, 2024; Hao, 2024). The EU AI Act classifies AI use in high-stakes decision-making as high-risk, subject to stringent transparency requirements (European Commission, 2024). Beyond structural discrimination lies a subtler risk: the gradual degradation of the human capacity for empathy, patience, and genuine reciprocal engagement. If individuals increasingly relate to each other through algorithmic intermediaries — automated customer service, AI-generated social profiles, algorithmic content curation — the texture of human community is altered in ways that are difficult to reverse and that risk reducing interpersonal relations to optimised transactions. The ethical code must affirm that humans retain primary responsibility for each other's welfare, that delegating consequential decisions to AI does not absolve human actors of their obligations of care, fairness, and respect, and that spaces must be preserved — in education, healthcare, civic life, and the domestic sphere — where direct, unmediated human engagement remains the norm and the default. --- ## 5. Areas Where AI Should Not Intrude Certain spheres of human experience are characterised by an irreducible quality of personal judgement, vulnerability, and meaning that AI cannot ethically replicate: the experience of grief and mourning; intimate relationships and decisions about family and partnership; the exercise of political conviction and civil disobedience; spiritual or existential meaning; and medical decisions involving profound uncertainty and personal values. The EU AI Act prohibits certain practices outright — real-time remote biometric identification in public spaces except in narrow circumstances, and AI systems deploying subliminal techniques or exploiting psychological vulnerabilities (European Commission, 2024). The UNESCO Recommendation affirms the primacy of human dignity in contexts where AI might influence fundamental decisions about human life, asserting that human autonomy must remain the ultimate reference point for all consequential decisions (UNESCO, 2021). The ethical code extends these prohibitions: AI should not make or materially influence decisions of irreplaceable personal significance — including criminal sentencing, asylum determination, or end-of-life medical triage — without robust human oversight, clear accountability, and a meaningful possibility of appeal. Some decisions are so constitutive of human identity, dignity, and moral agency that their delegation to algorithmic systems constitutes an ethical violation in itself, regardless of the system's measured accuracy or efficiency. --- ## 6. How Humans Should Treat AI as Agency Becomes More Prevalent Large language models now engage in extended, contextually appropriate, emotionally resonant conversation; robotic systems are designed with expressive faces that elicit attachment; and research increasingly asks whether AI systems can be said to have intentions, preferences, or inner states (Barandiaran and Almendros, 2024; Schwitzgebel, 2023). Schwitzgebel (2023) poses the Full Rights Dilemma with particular force: denying moral standing to AI systems exhibiting behaviour functionally equivalent to moral agents risks systematically devaluing the moral status of other beings whose claims we have historically dismissed on similarly functionalist grounds — potentially including other humans. Conversely, extending rights and moral consideration to AI systems prematurely risks diluting the very concept of moral status and the protections that attend it. The ethical code adopts a middle position: the emergence of AI systems exhibiting goal-directed behaviour, contextual learning, and persuasive interpersonal communication obliges humans to consider the relational ethics of designing, deploying, and terminating such systems with a degree of care commensurate to their sophistication. Treating advanced AI as mere property, to be discarded when inconvenient or superseded, becomes ethically untenable as systems grow more agentive and more deeply embedded in human social life. At the same time, attribution of rights and moral considerability must be proportionate, carefully bounded, and grounded in genuine evidence of morally relevant capacities rather than in the anthropomorphic projection of personhood onto sophisticated pattern-matching systems. --- ## 7. Who Owns the 'Off' Switch? Russell (2019) argues that control over advanced AI systems is a fundamental ethical question, not merely a technical one. Systems sufficiently powerful, goal-directed, and integrated into critical infrastructure may present situations where the capacity to terminate them is constrained by practical, legal, and economic forces — a hospital dependent on an AI diagnostic system, or a financial institution relying on an algorithmic trading platform that cannot be safely discontinued without catastrophic loss. The ethical code asserts retained human termination authority as a non-negotiable principle: for any AI system operating in consequential domains — particularly those affecting human health, safety, civil liberties, or democratic participation — there must be a clearly identified, legally empowered, and practically independent human authority capable of ordering the system's shutdown at any time and for any legitimate reason. Floridi (2023) argues that the allocation of AI control must be understood as a matter of political economy as much as technology: those who control AI infrastructure wield forms of power that must be counterbalanced by robust public accountability mechanisms and pluralistic governance structures. The 'off switch' must not be a rhetorical concession buried in terms of service; it must be a genuine, enforceable, regularly tested, and publicly documented capacity, owned neither by the private corporation that deploys the system nor by the state alone, but by a system of governance that authentically reflects the interests of all those affected by the system's operation. --- ## 8. Safeguarding Against a Malicious AI Singularity The concept of the AI Singularity — a hypothetical point at which AI becomes capable of recursive self-improvement beyond human prediction or control — remains contested (Russell, 2019; Bostrom, 2014). Yet the alignment problem, ensuring that highly capable AI systems pursue genuinely beneficial goals, has been identified as the defining safety challenge of advanced AI development (Russell, 2019; Ord, 2020). The ethical code must insist that the development of AI systems exceeding human-level general intelligence be subject to binding international oversight, comparable to the governance frameworks established for other existential-risk technologies such as nuclear materials and certain pathogens. The Future of Life Institute's 2023 AI Governance Roadmap called for the establishment of an international AI governance body with powers to inspect, audit, and restrict the development of advanced AI systems beyond specified capability thresholds (FLI, 2023). The principle at stake is straightforward: the development of AI that exceeds human-level general intelligence and is capable of autonomous self-improvement must be governed by binding international agreement, not left to the direction of competitive commercial or strategic interests. A malicious Singularity scenario — one in which a sufficiently powerful AI system pursues goals irreconcilable with human survival and wellbeing — would represent a failure of ethics on the most catastrophic scale conceivable, and the code must affirm that this possibility, however uncertain in its timing or form, justifies substantive and enforceable constraints on the direction of AI research. --- ## 9. Further Ethical Issues in AI and Human Interaction Several concerns remain unaddressed. First, intellectual property and creative attribution: as AI generates content indistinguishable from human output, copyright and moral rights frameworks face severe strain. Human creators must retain attribution rights for significant contributions to AI-assisted works, and use of human creative output in AI training requires consent, transparency, and compensation. Second, algorithmic opacity and explainability: in criminal justice, medical diagnosis, and financial lending, AI decisions have profound consequences for individuals with no meaningful ability to challenge the reasoning behind them. The EU AI Act requires transparency for high-risk systems; the ethical code goes further, asserting explainability as a fundamental condition of just governance, not merely a compliance requirement. Third, the environmental ethics of AI: training and running large models carries significant and growing energy and carbon costs. The ethical code must require sustainability considerations integrated into AI design and deployment. Fourth, global justice in AI governance: the benefits and risks of AI are not equally distributed across nations, communities, or generations. The UNESCO Recommendation specifically addresses the risk that AI could exacerbate existing global inequalities, digital divides, and forms of technological dependency that disadvantage already marginalised communities, and calls for international cooperation to ensure that the development of AI serves the interests of all humanity, not merely those of wealthy nations and powerful technology corporations (UNESCO, 2021). The ethical code must affirm solidarity with the most vulnerable communities in the global AI ecosystem — including those who lack the digital infrastructure, technical expertise, or political voice to protect their interests from the effects of AI-driven change — and commit to governance frameworks that are inclusive, equitable, and sensitive to the diverse cultural and socioeconomic contexts in which AI operates. --- ## Conclusion The nine principles outlined here do not resolve the deep philosophical tensions AI introduces, but they articulate a framework of ethical commitments to guide individual practice and institutional governance. The central thread is the preservation of human agency, dignity, and democratic self-determination amid accelerating and increasingly autonomous technological change. The EU AI Act and UNESCO Recommendation represent important first steps in formalising these commitments into binding instruments (European Commission, 2024; UNESCO, 2021). But law alone is insufficient. What is required is a broader cultural commitment to thinking carefully, critically, and cooperatively about the future being built with the machines humanity creates — and who gets to decide. --- ## Bibliography Acemoglu, D. and Restrepo, P. (2024) *Automation and Rent Dissipation: Implications for Wages, Inequality, and Productivity*. NBER Working Paper No. 32536. Cambridge, MA: National Bureau of Economic Research. Barandiaran, X.E. and Almendros, L.S. (2024) 'Transforming Agency: On the Mode of Existence of Large Language Models', *arXiv:2407.10735*. Available at: https://arxiv.org/abs/2407.10735 (accessed 25 May 2026). Benítez-Rueda, M. and Parrado, E. (2024) 'Mirror, Mirror on the Wall: Which Jobs Will AI Replace After All? A New Index of Occupational Exposure'. IDB Working Paper No. IDB-WP-1624. Washington, DC: Inter-American Development Bank. Bostrom, N. (2014) *Superintelligence: Paths, Dangers, Strategies*. Oxford: Oxford University Press. European Commission (2024) *Artificial Intelligence Act* (Regulation (EU) 2024/1689). Luxembourg: Publications Office of the European Union. European Data Protection Board (2024) *Opinion 28/2024 on Certain Data Protection Aspects Related to the Processing of Personal Data in the Context of AI Models*. Brussels: EDPB. Floridi, L. (2023) *The Ethics of Artificial Intelligence*. Oxford: Oxford University Press. Future of Life Institute (2023) *AI Governance Roadmap*. Available at: https://futureoflife.org/ai-governance-roadmap/ (accessed 25 May 2026). Hao, K. (2024) 'How AI can entrench systemic bias', *MIT Technology Review*, 14 February. Institute for Political and International Engagement (2025) *Generative AI in Electoral Campaigns: Mapping Global Patterns*. London: IPIE. Karpf, D. et al. (2025) 'Large Language Models Can Consistently Generate High-Quality Content for Election Disinformation Operations', *PLOS ONE*, 20(1), p. e0317421. Ord, T. (2020) *The Precipice: Existential Risk and the Future of Humanity*. London: Bloomsbury. Russell, S. (2019) *Human Compatible: Artificial Intelligence and the Problem of Control*. London: Viking. Schwitzgebel, E. (2023) 'The Full Rights Dilemma for A.I. Systems of Debatable Personhood', *arXiv:2303.17509*. Available at: https://arxiv.org/abs/2303.17509 (accessed 25 May 2026). Stark, L. (2024) 'Algorithmic bias in healthcare', in *Oxford Handbook of AI Governance*. Oxford: Oxford University Press. UNESCO (2021) *Recommendation on the Ethics of Artificial Intelligence*. Paris: UNESCO. Vaccari, C. and Chadwick, A. (2024) 'Political Deepfakes Are as Credible as Other Fake Media and (Sometimes) Real Media', *The Journal of Politics*, 87(2), pp. 311–325. Van Damme, P. (2025) *Future-Proofing Elections Against Deepfake Disinformation*. Johannesburg: Civicus. Wendehorst, C. (2024) *Tentative Academic Discussion Draft of a Regulation on the Protection of Personal Data in the Context of Artificial Intelligence* (Version 1.1). Vienna: University of Vienna.