- 22/12/2025
- Category: Commentaries
Author: Luthfi Baihaqi Riziq
Editor: Achmed Faiz Yudha Siregar
Introduction
Facing existential threats at the dawn of artificial intelligence and its rapid progress, governments and international bodies attempt to regulate and limit its production and use to certain standards. These framework documents appeal to ethical principles to induce legitimacy.[1] However, implementing a truly just regulation of AI for the benefit of humanity does not follow from adhering to principles. This is called the principles/practice gap in AI regulations.
Principles and Ambiguity
Approaching the AI issue with a universal-to-particular stance has been popular among the many entities developing research and regulatory frameworks. This seems intuitive and straightforward. A working government would need to anchor their actions to something; ethical principles are seemingly good choices. After all, we would need to grasp things we can agree on first before even thinking about operationalizing our efforts.
| Country/Body | Document | Principles Listed |
| UNESCO | Recommendation on the Ethics of Artificial Intelligence[2] | Ten core principles: Proportionality and do no harmSafety and securityRight to privacy and data protectionMulti-stakeholder and adaptive governance, and collaborationResponsibility and accountabilityTransparency and explainabilityHuman oversight and determinationSustainabilityAwareness and literacyFairness and non-discrimination |
| OECD / GPAI | OECD AI Principles[3] | Values-based principles: Inclusive growth, sustainable development, and well-beingRespect for the rule of law, human rights, and democratic values, including fairness and privacyTransparency and explainabilityRobustness, security, and safetyAccountability |
| China | Ethical Norms for Next Generation Artificial Intelligence 2021 (2021 AI Code of Ethics)[4] | Key principles: Enhancing human well-beingPromoting fairness and justiceProtecting privacy and safetyEnsuring controllability and reliabilityEnhancing accountabilityImproving ethical literacy |
| Australia | Australia’s AI Ethics Principles[5] | Human, societal, and environmental well-beingHuman-centred valuesFairnessPrivacy protection and securityReliability and safetyTransparency and explainabilityContestabilityAccountability |
| Indonesia | Communication and Informatics Ministerial Circular on AI Ethics[6] | InclusivityHumanitySecurityAccessibilityTransparencyCredibility and accountabilityPrivate data protectionSustainable development and environmentIntellectual property |
The above table specifies several examples of principles that are presented in these frameworks. They concisely describe the desired values for AI limitations, giving prompt sense to government or industrial actors about abstract ideals. These principle-based documents are an important and necessary step in the evolution of AI governance.[7] However, it only acts as a momentary overview of expectations.
The main problem with principlism, i.e. relying on principles, in AI governance is that it provides no guidance to AI practitioners on how to design or implement algorithms within the set ethical boundaries.[8] Yet, as constitutions give no instructions on how to establish laws, so too ethical frameworks give no real directive on the technicalities of regulation. It was never meant to bind. In other words, it serves no practical function.
Mentioning values like transparency and sustainability, for example, does not automatically grant urgency for government or industrial actors to be transparent and sustainable.[9] The inherent vagueness and ambiguity of the principles make it commonsensical and ignorable at best or give way to exploits at worst. In these documents, AI ethics is only appropriated to signify moral high ground and meaningful participation.
Even when UNESCO follows the principles with “actionable policies”, the principles still do not translate to practice. For one, ideals like justice, sustainability, and proportionality are contested social concepts—changing between individuals, places, and cultures. Procedural assessments of ethical impacts based on the principles may attempt to measure using the most extreme imaginable threshold,[10] but it does not matter in AI development or use. Even explicitly telling developers to consider ethical codes produces unchanging attitudes; they prioritize the true fast and frugal heuristics: economic or technical factors.[11]
Through the Looking-Glass
An alternative to AI ethics based on principles is one that lowers the guidance to the level of design and technical specifications.[12] Instead of listing value prompts, this might look like a toolkit with sets of “how” questions, expandable but urging systematic steps to be taken. As an example, Digital Catapult’s AI Ethics Framework uses a combination of what and how questions to address the design of an AI product and its oversight mechanisms.[13]
Identifying tools and methods that might help in design situations is important to set clear boundaries in AI design from start to finish.[14] It should include maps of which party can be held accountable, which should take action when, and what language to use. AI governance should be pragmatic, in that it should strive to be operationalizable instead of ideal. Preparing how to regulate trumps asking what sets the end goal. Instead of proposing universally recognized principles, meeting AI practitioners halfway could produce a more relevant ethical and regulatory mechanism.
The limitation of this is that it will leave some factors out in favor of implementing mapped procedures.[15] As such, it should be noted that toolkits should be regularly updated, even faster than regulatory frameworks can keep up. It may elaborate principles we have reached consensus on, or it might discard them entirely. However, it should not be implemented top-down as principles would, but rather bottom-up or middle-out—enacted by developing or impacted organizations, expanding its relevancy from specific use cases to cross-organizational adaptation.[16]
Currently, the Indonesian government has not implemented a practitioner-based guide or bill that can help the design and adoption of AI. The Ministerial Circular mentioned above ushered principles, while the National Roadmap for AI[17] highlights a general direction of policies up to 2029. AI ethics is still reduced to principles and the governance plans to develop, adopt, and adapt are not necessarily implied by the mentioning of principles. Instead, AI ethics toolkits to expand know-how surrounding AI safeguarding can be designed specifically for and by way of experience from developers, infrastructure experts, civil organizations/communities, and certain vulnerable impacted groups. These will help supply a readily available knowledge base as practitioners try to harness optimal benefits from AI tools for human development and mitigate its potential harms.
Conclusion
The not-yet actionable AI ethical principles should be complemented with meso-level guidance to AI practitioners on every aspect of concern. Design to distribution to use always needs constant feedback from all working practitioners and users. As such, a multi-disciplinary ethical audit board, consisting partially of AI practitioners, is in order to design and problematize both the practice and the ethics actively.
[1] For more discussion about legitimacy of AI ethics, see Woods, D. (2025). Stag hunt in the digital wilds: Legitimizing global AI governance amidst diverse terrains. Fudan Journal of the Humanities and Social Sciences. https://doi.org/10.1007/s40647-025-00438-3
[2] UNESCO. (2022). Recommendation on the ethics of artificial intelligence. https://unesdoc.unesco.org/ark:/48223/pf0000381137.locale=en
[3] First adopted 2019, then amended 2024. See OECD. (2019). OECD AI principles overview. https://oecd.ai/en/ai-principles
[4] Yang, S., Fung, C., & Zhou, B. (2025, January 20). AI ethics: Overview (China). China Law Vision. https://www.chinalawvision.com/2025/01/digital-economy-ai/ai-ethics-overview-china/. The original Mandarin can be accessed here: https://www.most.gov.cn/kjbgz/202109/t20210926_177063.html
[5] Australian Government. (2019, November 7). Australia’s AI ethics principles. Australian Department of Industry, Science and Resources. https://www.industry.gov.au/publications/australias-ai-ethics-principles
[6] The ministerial circular can be accessed here: https://jdih.komdigi.go.id/produk_hukum/view/id/883/t/surat+edaran+menteri+komunikasi+dan+informatika+nomor+9+tahun+2023
[7] See Raab, C. D. (2020). Information privacy, impact assessment, and the place of ethics. Computer Law & Security Review, 37, 105404. https://doi.org/10.1016/j.clsr.2020.105404
[8] Orr, W., & Davis, J. L. (2020). Attributions of ethical responsibility by Artificial Intelligence practitioners. Information, Communication & Society, 23(5), 719-735. https://doi.org/10.1080/1369118X.2020.1713842
[9] Munn, L. The uselessness of AI ethics. AI and Ethics, 3, 869-877. https://doi.org/10.1007/s43681-022-00209-w
[10] UNESCO. (2023). Ethical impact assessment: A tool of the Recommendation on the Ethics of Artificial Intelligence. https://unesdoc.unesco.org/ark:/48223/pf0000386276
[11] Hagendorff, T. (2020). The ethics of AI ethics: An evaluation of guidelines. Minds and Machines, 30, 99-120. https://doi.org/10.1007/s11023-020-09517-8
[12] Morley, J., Elhalal, A., Garcia, F., Kinsey, L., Mokander, J., & Floridi, L. (2021). Ethics as a service: A pragmatic operationalisation of AI ethics. Minds & Machines, 31, 239-256. https://doi.org/10.1007/s11023-021-09563-w
[13] Digital Catapult. (2018). Ethics framework. https://www.digicatapult.org.uk/wp-content/uploads/2023/06/DC_AI_Ethics_Framework-2021.pdf
[14] Morley, J., Floridi, L., Kinsey, L., & Elhalal, A. (2020). From what to how: An initial review of publicly available AI ethics tools, methods and research to translate principles into practices. Science and Engineering Ethics, 26, 2141-2168. https://doi.org/10.1007/s11948-019-00165-5
[15] Every framework and toolkit can be limited to certain situations. For more discussion, see Qiang, V., Rhim, J., & Moon, A. (2024). No such thing as one-size-fits-all in AI ethics frameworks: A comparative case study. AI & Society, 39, 1975-1994. https://doi.org/10.1007/s00146-023-01653-w
[16] See Wong, R. Y., Madaio, M. A., & Merrill, N. (2023). Seeing like a toolkit: How toolkits envision the work of AI ethics. Proceedings of the ACM on Human-Computer Interaction, 7(CSCW1). https://doi.org/10.1145/3579621
[17] The National Roadmap for AI draft was released August 2025 and is available here: https://s.komdigi.go.id/KonsultasiPublik_KA
