Skip to content
Data Protection

Artificial Intelligence and Automated Decision-Making under Ethiopian Data Protection Law

By Haymanot & Advocates 18 November 2026 18 min read

Computer code on a screen

INTRODUCTION

Artificial intelligence (AI) has become one of the most transformative technologies of our time. Once confined to research labs or tech giants, AI now powers tools and services that shape daily life from predictive search engines and personalized recommendations to complex systems that assist in healthcare diagnostics, urban planning, legal and financial risk analysis.[1] At its simplest, AI refers to machines or systems that can perform tasks that typically require human intelligence, such as recognizing patterns, making decisions, learning from data, and adapting to changing conditions.[2] This capability allows AI to enhance productivity, optimize service delivery, and offer data-driven solutions to persistent development challenges, particularly in resource-constrained environments.

However, AI’s ability to process large volumes of data introduces not only technical and economic opportunities but also critical challenges especially when it comes to the protection of personal data. Modern AI systems depend heavily on data at every stage of their lifecycle, from training to deployment. In many instances, this data includes information that directly or indirectly identify individuals. Whether collected from users themselves or from secondary sources, this personal data plays a foundational role in how AI systems learn, make predictions, and evolve over time. As AI systems grow increasingly data-dependent, their functioning becomes inseparable from the legal and ethical obligations surrounding data governance. Personal data especially when aggregated or used in automated decision-making becomes more than just a technical input. It implicates fundamental rights, including privacy, consent, and non-discrimination. These concerns become more pronounced when AI systems function as “black boxes,” generating outcomes that are difficult to interpret or challenge. In such instances, AI-driven decisions may result in unintended discrimination, undermine informed consent, and erode accountability.

Ethiopia is beginning to position itself within this global movement. As part of its efforts to embrace digital transformation, the government endorsed the National Artificial Intelligence Policy (the “Policy”) in June 2024.[3] This Policy sets out a strategic vision to encourage AI research, foster innovation, and support the application of AI across sectors such as health, agriculture, education, and public administration.[4] This initiative is reinforced by the Digital Ethiopia 2025 Strategy, which aims to strengthen digital infrastructure and modernize public services.[5] Recent studies on AI readiness in Ethiopia, however, highlight ongoing structural challenges such as limited private sector participation, data fragmentation, underdeveloped digital infrastructure, and the absence of high-performance computing capabilities. The government remains the primary actor in advancing AI, although promising use cases are emerging in areas like agriculture, healthcare, and language technologies. However, As AI applications begin to scale in these sectors, their reliance on personal data becomes increasingly pronounced. Whether applied in predictive health diagnostics, smart farming platforms, or digital identity systems, AI technologies frequently process data that directly or indirectly identifies individuals. This interdependence between AI functionality and personal data raises significant governance concerns, especially in a context where legal and institutional safeguards are still taking shape. [6]

Although the Personal Data Protection Proclamation No. 1321/2024 (PDPP) outlines important principles such as data minimization, purpose limitation, and provisions related to automated decision-making it does not yet provide a fully developed framework for managing the unique risks posed by AI. Issues like algorithmic bias, lack of transparency, and the challenge of tracing accountability in automated processes remain insufficiently addressed. As Ethiopia continues to build its digital future, the integration of AI with robust personal data protection mechanisms will be central to ensuring responsible innovation and safeguarding individual rights. This article, as part of our broader series on personal data protection, explores key considerations in AI governance, the regulation of AI-driven data processing, and the critical importance of transparency and accountability in automated decision-making within the Ethiopian context.

GLOBAL FRAMEWORKS FOR ETHICAL AI GOVERNANCE

As AI continues to advance, its growing complexities have prompted international organizations and national stakeholders to develop frameworks that ensure ethical and responsible AI development.[7] While these frameworks differ in scope and implementation, they share fundamental principles: transparency in algorithms and decision-making, fairness to prevent bias and discrimination, harm minimization (non-maleficence), clear accountability, and strong data privacy safeguards. Among the most notable global initiatives in UNESCO’s Recommendation on the Ethics of AI, which promotes a human-centered framework grounded in inclusivity, sustainability, and the protection of human rights. Endorsed by 193 countries, including Ethiopia, this recommendation marks a significant milestone in the global governance of AI..[8] The OECD AI Principles similarly emphasize transparency, accountability, and adherence to the rule of law as essential foundations for building public trust in AI.[9] The G7 Hiroshima AI Process introduced a voluntary Code of Conduct for developers, focusing on enhancing safety and mitigating risks linked to advanced AI systems. The G20 AI Principles aligned with those of the OECD, further highlight the importance of international cooperation and inclusive growth.[10] Most recently, the United Nations General Assembly adopted its first resolution on AI,[11] calling for the development of safe, secure, and trustworthy AI systems that support sustainable development goals and help bridge the global digital divide.

Apart from these soft law efforts, the European Union has adopted a more binding regulatory approach through the AI Act. This legislation establishes a risk-based framework that imposes strict requirements on high-risk AI applications and prohibits those deemed unacceptable. In this regard, the Act underscores the principles of transparency, safety, and accountability, particularly with respect to AI systems that process personal data. It also references the General Data Protection Regulation (GDPR) in over 30 provisions, emphasizing the integral role of data protection in AI governance. A key principle shared by both the AI Act and the GDPR is Privacy-by-Design, which mandates that privacy safeguards be embedded into the system architecture from the earliest stages, rather than added as an afterthought. This approach ensures that privacy is treated as a foundational element of system design. However, applying Privacy-by-Design in the AI context presents unique challenges. AI systems frequently involve large-scale and dynamic data processing, often leading to complex, opaque decision-making processes. These characteristics make it difficult to foresee and control privacy risks at every stage of development and deployment.

ETHIOPIA’S APPROACH TO AI GOVERNANCE

In line with these global trends, Ethiopia has also taken a proactive step to build a foundation for AI governance The country’s National Artificial Intelligence Policy, endorsed in June 2024, outlines a strategic vision for leveraging AI in support of national development priorities. The Policy adopts a context-sensitive approach, recognizing the transformative potential of AI technologies while acknowledging the associated risks particularly those related to personal data governance. The Policy emphasizes the creation of an enabling infrastructure, proposing strategies to support AI development across key sectors such as healthcare, water and energy, e-government, education, mobility and urban planning, financial inclusion, national defense, agriculture, tourism, creative industries, and manufacturing. These sectors are identified as areas where AI can have the most significant impact given Ethiopia’s socio-economic context.[12]

Despite this strategic focus, Ethiopia’s AI development is still constrained by practical limitations. A 2025 National AI readiness report highlights several pressing challenges, including limited access to high-quality datasets particularly in underrepresented local languages and a shortage of skilled professionals trained in machine learning, data science, and ethics. While some startups and institutions are engaged in promising innovation, they often face hurdles such as inadequate funding, underdeveloped digital infrastructure, and restricted access to AI computing resources. These limitations pose barriers to scaling AI solutions beyond pilot phases and risk deepening existing inequalities in digital access and participation.[13]

From data governance perspective, the Policy underscores the need for robust evaluation and oversight mechanisms to ensure the responsible and ethical deployment of AI systems. It introduces a risk-adjusted governance model, which categorizes sectors based on their sensitivity, with high-risk areas such as government decision-making, healthcare, and national security subject to enhanced scrutiny. In these domains, where AI systems are likely to process large volumes of sensitive personal data, the Policy insists on close alignment with existing non-automated procedures to preserve public trust and avoid unintended consequences. The Policy also calls for ongoing assessments of AI tools, requiring both compliance checks and cost-of-error analyses, especially where misuse of data could result in harm to individuals or public services.

In this context, data protection is not merely a complementary concern but a core component of responsible AI usage. The power of AI lies in its ability to analyze and learn from vast amounts of data, much of which may involve sensitive or personally identifiable information. As such, ensuring that this data is collected, processed, and stored in a manner that respects privacy, consent, fairness, and accountability is critical to maintaining ethical standards in AI deployment. The AI Policy acknowledges this interdependence and stresses that AI development must be guided by core principles of transparency, human rights, and data protection.

Nevertheless, Ethiopia has made significant strides with the enactment of the PDPP. The PDPP establishes a framework for data governance, delineating clear rights for data subjects and defining the obligations of data controllers and processors. Notably, PDPP enshrines critical data protection principles such as purpose limitation, data minimization, storage limitation, data sovereignty, and conditions for international data transfers which could be essential for ensuring the ethical and secure handling of data in AI applications. Though these principles may not fully capture the complexities of AI governance, their applicability could be expanded to address emerging challenges in AI-driven data processing.

4. INSTITUTIONAL FRAMEWORK AND REGULATORY BODIES

In terms of regulatory arrangements, Ethiopia has established a multi-layered institutional and framework to govern the development and use of artificial intelligence. According to Proclamation No. 1263/2021 on the Powers and Duties of the Executive Organs, the Ministry of Innovation and Technology (MinT) is mandated to initiate policies, strategies, laws, and programs that promote the sustainable development of emerging technologies.[14] To give practical effect to this mandate, among other, the Ethiopian Artificial Intelligence Institute (the “Institute”) was set up through Council of Ministers Regulation No. 510/2022 which is answerable to MinT. The Institute is tasked with drafting and overseeing the implementation of national AI policies and legal instruments. It also plays a strategic role in collecting and analyzing data to support national security decision-making, assisting security agencies, and ensuring that AI technologies contribute to economic and social development.

Although the Institute’s activities may incidentally involve personal data particularly in the development and deployment of AI systems regulation of such data processing falls primarily under the Ethiopian Communications Authority (ECA), as designated by the PDPP. While the PDPP does not expressly refer to AI, its general provisions on automated decision-making and data protection principles apply to AI-driven personal data processing. Accordingly, the ECA can play a central role in overseeing these practices to ensure they are carried out in a transparent, lawful, and rights-respecting manner. One point worth noting in this regard is that sector-specific authorities may also become involved in the deployment of AI systems within their respective mandates. In addition to central institutions, these regulators are increasingly shaping AI governance through domain-specific initiatives. For example, the Ethiopian Food and Drug Authority issued the 2022 Guidelines for Software as a Medical Device and AI-Based Medical Devices, setting standards for the use of AI in healthcare. Similarly, other bodies such as the National Bank of Ethiopia, the Ministry of Education, and the Ministry of Agriculture are anticipated to develop and implement sectoral frameworks that promote the ethical, safe, and effective use of AI in line with public interest and national development goals. While their primary focus may not be on data protection, their involvement in AI deployment may indirectly intersect with personal data processing, making coordination with the ECA increasingly important. .

THE PDPP IN THE AI CONTEXT

AI systems process personal data at different stages, including during training, prediction, and real-world applications. In the training phase, datasets often include personal information such as age, occupation, health records, or financial history. Even when such data is anonymized where personal identifiers are removed in a way that makes re-identification theoretically impossible or pseudonymized where identifiers are replaced with fictitious markers but can be linked back using additional information, there remains a risk of re-identification if the dataset includes enough specific or unique attributes. After training, AI models may inadvertently expose personal information through techniques like model inversion , which can reconstruct parts of the original input data, for example recovering a blurred face from a facial recognition model, or membership inference attacks, which can reveal whether an individual’s data was part of the training dataset. In the prediction phase, AI systems continue processing personal data, such as analyzing data from smartwatches or tracking user behavior based on location. Often, the data collected exceeds what is necessary, raising concerns about over-collection and violating the principle of data minimization.[15] A notable case in this regard occurred on March 30, 2023, when the Italian Data Protection Authority ordered OpenAI to halt the use of ChatGPT for processing the personal data of Italian data subjects. The DPA cited a significant risk of non-compliance with the GDPR, particularly due to the use of personal data for ChatGPT’s training set without an adequate legal basis or a proper privacy notice. This ruling prompted OpenAI to take corrective measures.[16]

In the Ethiopian context, the PDPP applies whenever AI-driven systems process the personal data of natural persons, whether in whole or in part through automated means or as part of a filing system. Personal data, as defined under the PDPP, refers to any information that relates to an identified or identifiable natural person.[17] Therefore, whenever an AI system processes personal data whether during training, operation, or inference it falls under the scope of the PDPP, unless an exemption applies. While the PDPP does not explicitly mention AI-driven data processing, its principles and provisions are applicable and extend to regulate such activities when personal data is involved.

Core Principles of the PDPP in Relation to AI Systems

  • Lawfulness, Fairness, and Transparency:[18] One of the key challenges in AI-driven data processing is ensuring transparency in how personal data is used. While the PDPP mandates that data processing must be lawful, fair, and transparent, AI systems often operate with complex algorithms that may not be easily understood by the data subjects. This becomes even more problematic in cases like predictive analytics or recommendation systems, where the algorithms may combine data from different sources to make decisions in ways that are not immediately apparent to the users.[19]
  • Purpose Limitation:[20] The principle of purpose limitation in the PDPP ensures that personal data is collected only for explicit and legitimate purposes, and it cannot be further processed in ways that are incompatible with those purposes. However, AI systems often require the reuse of data that was initially collected for one purpose and applied to entirely different uses. For instance, data gathered for a specific purpose, such as user registration for an online platform, might later be repurposed for AI-driven marketing or personalized recommendations. The reuse of data for purposes not anticipated at the time of collection can lead to function creep, where the data is used in ways that exceed the scope of the original purpose.
  • Storage Limitation:[21] The principle of storage limitation mandates that personal data should not be kept longer than necessary for the purposes for which it was collected. However, AI systems, especially those that rely on large datasets for training, often face challenges in meeting this requirement. Since machine learning models continuously evolve, they may need to retain older data for ongoing refinement and development, making it harder to ensure timely erasure of data.

6. AUTOMATED DECISION MAKING AND PDPP

Automated decision-making refers to a process where decisions affecting individuals are made by automated systems. These systems often use algorithms, AI, and data analytics to assess and make decisions about individuals based on the data available. In many cases, automated decision-making is used for tasks such as credit scoring, hiring processes, loan approvals, and targeted marketing.[22] In contrast, general data processing typically involves a significant degree of human input data may be collected, organized, or analyzed by automated tools, but the final decision is usually made or interpreted by a human. What sets automated decision-making apart is that the system itself determines the outcome, often with little to no human review, which raises unique concerns around fairness, transparency, and accountability.[23]For example, an AI algorithm used in recruitment may unintentionally favor certain demographics over others, leading to discrimination or bias in hiring decisions. As such, it becomes essential to regulate automated decision-making to protect the legal rights and freedoms of individuals.

Under the PDPP individuals have the right not to be subject to decisions made solely through automated processing, including profiling, when those decisions produce legal effects or similarly significant impacts.[24] Although the PDPP does not provide detailed guidance on the application of this right, it is particularly relevant in contexts where decisions significantly affect someone’s rights or opportunities. Consider, for instance, a university receiving thousands of applications or an employer managing mass recruitment. To handle such volumes efficiently, institutions may use automated tools that scan applications for specific keywords or qualifications sometimes resulting in rejections without any human review. While this increases efficiency, it raises serious concerns when the decision has a substantial impact, such as denial of admission or employment.

The PDPP does not clearly define what constitutes “human intervention,” but the law appears to intend more than a symbolic review. It suggests that the intervention should be meaningful and undertaken by someone with both the authority and understanding to assess the system’s decisions and, if necessary, override them. This kind of review must go beyond merely approving the automated result and should ensure that individual circumstances are fairly considered.

The Proclamation also grants data subjects the right to express their views regarding automated decisions. Although it does not specify how this should be exercised, it can be reasonably interpreted to include the right to contest the outcome and question the logic or criteria used by the system particularly in cases where individuals feel the process ignored relevant context or made inaccurate assumptions based on limited data. Furthermore, the PDPP explicitly prohibits automated decision-making that evaluates personal aspects based on sensitive personal data such as ethnicity, health status, or biometric identifiers further emphasizing the law’s cautious approach to sensitive profiling.

This general rule on automated decision-making also allows for certain exceptions such as when the decision is necessary for entering into or performing a contract, is authorized by law, or is based on the data subject’s explicit consent. Even in such cases, however, data controllers are required to meet specific conditions. For instance, when relying on consent or contractual necessity, they must clearly and proactively inform data subjects about the existence of the automated processing and its potential consequences. Moreover, the consent must align with the PDPP’s standards: it must be freely given, specific, informed, and unambiguous.

6.1 TRANSPARENCY AND ACCOUNTABILITY IN AI DECISION-MAKING

The issue of transparency in AI systems has become a significant concern, as AI models increasingly influence important aspects of people’s lives. Many AI systems operate as “black boxes,” where the internal workings and decision-making processes are not readily accessible or understandable to those affected by the outcomes.[25]

The PDPP attempts to address this transparency concern by requiring data controllers to inform individuals about the existence of automated decision-making processes, including profiling.[26] While the law doesn’t specify the level of detail required, it can be inferred that the focus is not on intricate technical explanations of algorithms, but rather on details regarding how these systems are practically utilized and how they impact individuals. This might include informing data subjects about the types of data used in the decision-making process, the rationale behind selecting those specific factors, how the data is processed and weighted, any statistical or models applied, and the relevance of these factors to the decision being made about the data subject. To strengthen individual protections in such contexts, the PDPP also sets out clear enforcement mechanisms. It establishes criminal penalties for those who violate core data subject rights. Specifically, anyone who does not respect the right against automated decisions shall be punished with imprisonment for a period ranging from three to five years or fined between 100,000 and 200,000 birr, or both.[27]

Nonetheless, achieving transparency may be challenging in practice even before accounting for the added complexity introduced by AI systems. Modern AI, especially those powered by deep learning, depends on highly intricate models trained on vast datasets. This makes it extremely difficult to trace how a particular outcome is generated, especially as these models continuously learn and evolve. The inherent opacity of such systems, combined with potential biases in the training data, often makes it hard for individuals to understand or contest the decisions made particularly when those decisions have serious legal or personal implications.[28]

7. FUTURE OUTLOOK AND RECOMMENDATIONS

While PDPP introduces important principles on automated decision-making and profiling, it may not be sufficient to fully address the evolving risks posed by AI. As technology progresses, the PDPP should be expanded or supplemented by specific AI governance laws or regulations. These should offer clearer definitions, introduce safeguards for high-risk AI uses, and require tools like algorithmic audits and impact assessments. In this regard, Ethiopia may draw lessons from global approaches such as the EU AI Act, which complements the GDPR and cross-references it in multiple provisions to ensure strong protection of individuals’ data rights in the context of AI deployment. In the meantime, Ethiopia should consider requiring technical security risk assessments for AI systems, especially in sensitive areas such as finance, employment, or public services. These assessments should evaluate potential harms such as bias, misuse of personal data, or security flaws. To ensure effective enforcement, regulatory institutions should be strengthened with the technical and legal capacity to oversee AI-related risks. As AI continues to evolve, Ethiopia’s legal framework should remain flexible and responsive to emerging challenges.

  1. Bryant, K. (2023, December 13). How AI Is Impacting Society And Shaping The Future. Retrieved April 3, 2025, from Forbes website: https://www.forbes.com/sites/kalinabryant/2023/12/13/how-ai-is-impacting-society-and-shaping-the-future/ ↑

  2. Contributors to Wikimedia projects. (2025b, April 1). Artificial intelligence. Retrieved April 3, 2025, from Wikipedia website: https://en.wikipedia.org/wiki/Artificial_intelligence ↑

  3. Admin. (2024, June 27). The council of ministers unanimously decide to implement the national artificial intelligence policy – Artificial intelligence institute. Retrieved April 3, 2025, from Artificial Intelligence Institute – website: https://aii.et/the-council-of-ministers-unanimously-decide-to-implement-the-national-artificial-intelligence-policy/ ↑

  4. AI Policy ↑

  5. Digital Ethiopia 2025 – A Strategy for Ethiopia Inclusive Prosperity. Retrieved April 3, 2025, from https://www.lawethiopia.com/images/Policy_documents/Digital-Ethiopia-2025-Strategy-english.pdf ↑

  6. GSMA.. (2025, April 5), AI in Ethiopia: Promising use cases for development. Retrieved from https://www.gsma.com/solutions-and-impact/connectivity-for-good/mobile-for-development/wp-content/uploads/2025/04/AI-in-Ethiopia.pdf ↑

  7. Gatto, J. G. (2024, May 3). Responsible AI – Everyone is Talking About it But What Is It? Sheppard Mullin Richter & Hampton LLP. Retrieved April 4,2025, from https://www.lexology.com/library/detail.aspx?g=4b7d5f0a-2026-4c32-be9a-e29e561f317f ↑

  8. Recommendation on the ethics of artificial intelligence. (2024, September 26). Retrieved April 4, 2025, from UNESCO website: https://www.unesco.org/en/articles/recommendation-ethics-artificial-intelligence ↑

  9. AI principles. (n.d.). Retrieved April 4, 2025, from OECD website: https://www.oecd.org/en/topics/ai-principles.html ↑

  10. White & case LLP international law firm, global law practice. (n.d.). Retrieved April 4, 2025, from White & Case LLP website: https://www.whitecase.com/insight-our-thinking/ai-watch-global-regulatory-tracker-g7 ↑

  11. General Assembly adopts landmark resolution on artificial intelligence. (2024, March 21). Retrieved April 4, 2025, from UN News website: https://news.un.org/en/story/2024/03/1147831 ↑

  12. Supra note 5. Page 7 ↑

  13. Supra note 7 ↑

  14. Definition of Powers and Duties of the Executive Organs Proclamation No. 1263/2021. Article 29(1)(A) ↑

  15. How should we assess security and data minimisation in AI? (2024, December 17). Retrieved April 5, 2025, from ICO website: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/how-should-we-assess-security-and-data-minimisation-in-ai/ ↑

  16. Rahman-Jones, B. I. (2024, January 29). ChatGPT: Italy says OpenAI’s chatbot breaches data protection rules. Retrieved April 7, 2025, from BBC News website: https://www.bbc.com/news/technology-68128396 ↑

  17. PDPP, Article 2(2) ↑

  18. PDPP, Article 6(1) ↑

  19. Craig, L. (2024, October 10). Interpretability vs. explainability in AI and machine learning. TechTarget. Retrieved from https://www.techtarget.com/searchenterpriseai/feature/Interpretability-vs-explainability-in-AI-and-machine-learning ↑

  20. PDPP, Article 13. ↑

  21. PDPP, Article 15 ↑

  22. How should we assess security and data minimisation in AI? (2024b, December 17). Retrieved April 5, 2025, from ICO website: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/how-should-we-assess-security-and-data-minimisation-in-ai/ ↑

  23. Ibid ↑

  24. PDPP, Article 31 ↑

  25. Supra note 26. ↑

  26. PDPP, Article 24, m ↑

  27. PDPP, Article 64 (2)(d) ↑

  28. Transaparncy issue in deep learning models (March 17, 2025), Retrieved from https://www.restack.io/p/ai-interpretability-answer-transparency-issues-deep-learning-cat-ai ↑

This article is general information and not legal advice. Contact us about your specific situation.

Data Protection artificial intelligence data protection
H&A

Published by

Haymanot & Advocates

Legal update · Data Protection practice