AI Policy Archives – Collaboration on International ICT Policy for East and Southern Africa (CIPESA)

By Raylenne Kambua |

In August 2026, the Collaboration on International ICT Policy for East and Southern Africa (CIPESA) submitted a detailed set of recommendations to the Committee on the Draft Kenya Artificial Intelligence (AI) and Other Emerging Technologies Policy, 2026. The submission calls for closer alignment of Kenya’s policy with regional and international AI frameworks, alongside stronger protections for fundamental human rights.

The submission emphasises the need for Kenya to strike a balance between its aspirations for AI governance and strong safeguards that protect human rights, people, and democratic values. Without these, the nation risks developing systems that are innovative but exclusive, unfair, and harmful.

CIPESA argues that since Kenya is simultaneously advancing multiple AI-related processes, including a national AI strategy, a proposed AI Bill, and this draft AI policy, there is a need to align and harmonise the different proposed frameworks to ensure coherence and mitigate the risks of duplication and contradictory provisions.

For example, under section 3.5.1, the Policy commits the government to develop a dedicated AI and other Emerging Technologies Governance Act to provide the legal framework for its governance in Kenya, including the establishment, powers, and functions of the Council, without acknowledging that a similar Bill is already before the Senate.

Kenya’s policy direction is influenced by constitutional obligations as well as wider regional and international commitments that underscore the importance of human rights, accountability, transparency, and inclusiveness. According to CIPESA’s submission, Kenya should incorporate these principles into legally binding policy measures rather than just mentioning them.

The submission further encourages collaboration within the East African region and across the African continent, pointing to the value of shared standards, combined knowledge, and coordinated advocacy in strengthening governance outcomes.

CIPESA’s Navigating the Implications of AI in Kenya report also highlights that AI is reshaping digital participation, information access, and democracy in Kenya. In the absence of explicit protections and clear safeguards, AI systems can perpetuate discrimination, facilitate surveillance, violate people’s right to privacy, restrict freedom of expression, and undermine livelihoods. A rights-based approach that includes mandatory human rights impact assessments will ensure that potential harms are identified and mitigated before systems are deployed.

While the draft policy outlines institutional structures and governance ambitions, CIPESA argues that effective oversight will depend on institutional independence, clear powers, and meaningful accountability. The submission raises concerns about the proposed AI Council’s institutional independence and recommends giving it explicit authority to audit, obtain information, enforce compliance, and report directly to Parliament.

In automated systems, decision-making processes are often opaque and distributed across multiple actors. CIPESA therefore recommends clearly defining responsibilities and liability so that individuals harmed by AI systems have effective redress mechanisms.

Effective AI governance requires technical expertise, resources, and coordination across multiple agencies, yet many institutions in Kenya remain under-resourced. Therefore, proposed governance frameworks should be realistic about the state’s ability to implement and enforce stronger oversight mechanisms by investing in institutional capacity and talent retention.

According to CIPESA’s research, AI content moderation on major platforms is built largely for the Global North, with low-resource African languages. Many AI systems deployed in African contexts are trained on datasets that do not reflect local realities, leading to biased outcomes with direct implications for fairness, inclusion, and accuracy. This necessitates strong local data ecosystems and locally relevant content moderation systems and languages.

Kenya’s 2025 High Court ruling on the Worldcoin iris-scanning project affirmed the need for stronger data protection measures and integration with AI-specific legislation. According to CIPESA, incorporating pre-deployment oversight would transform AI governance from a reactive to a proactive model, particularly regarding sensitive biometric data.

AI systems are resource-intensive, consuming a lot of energy and requiring large amounts of water for data centre cooling. They also emit carbon and ultimately contribute to electronic waste. The submission recommends environmentally sustainable approaches, including independent third-party verification of environmental disclosures and publication of verified information in the public Registry.

The African Union AI Strategy identifies disinformation as a distinct risk. AI can influence public discourse by deciding which information is promoted, suppressed, or amplified. Disinformation, manipulation, targeted harassment, technology-facilitated gender-based violence (TFGBV), and AI-generated deepfakes can create particular risks in civic and democratic spaces, with disproportionate effects on women and other vulnerable groups. CIPESA proposes explicit recognition of these threats, implementation of gender impact assessments for high-risk systems, and stronger oversight of AI use in elections, political advertising, and content moderation.

Another recommendation is the inclusion of civil society representation at the steering committee, which is the top decision-making level. This is to ensure meaningful participation and alignment with the African Union AI Strategy and the UNESCO Recommendation on the Ethics of AI, which call for inclusive, multi-stakeholder involvement in AI governance, especially where major decisions are made.

Inclusion and public participation also require accessible language and processes that enable broader public engagement with what are often complex and technical issues, through investments in digital literacy and public awareness.

The submission further underscores the importance of labour rights and the often invisible workforce behind AI systems, many of whom work in unfavourable conditions. By highlighting the need for fair labour standards, protections, and recognition of data work, CIPESA also recommends including the workforce that sustains AI ecosystems in high-level policy discussions.

Explainability and transparency are essential to accountable AI governance. For people to trust AI systems, they need to understand how AI-driven decisions are made and contest results they believe to be unfair. To prevent AI systems from being treated as black boxes beyond public scrutiny, there must be clear documentation, disclosure standards, and rights to explanation. This builds accountability and trust, especially in high-risk sectors like public services, healthcare, and finance.

Finally, CIPESA highlights the necessity of continuous policy review and adaptation, emphasising that governance frameworks must remain adaptable and responsive as AI technologies evolve quickly. This includes establishing mechanisms for periodic review, stakeholder feedback, and iterative policy development to ensure that regulations remain relevant and effective over time.

CIPESA’s recommendations provide a mechanism to close the gap between ambition and accountability as Kenya works to finalise its AI policy. The decisions made at this point will influence not only the development, deployment, and application of AI but also the distribution of its benefits and risks.

Key recommendations from CIPESA:

Align Kenya’s AI policy and legislation to avoid duplication and conflicting provisions.
Make human rights and gender impact assessments mandatory for high-risk AI systems.
Strengthen the independence and powers of AI oversight institutions, including audit, enforcement, and redress.
Protect workers across the AI value chain, including data annotators and content moderators.
Require environmental accountability for AI, including disclosure and independent verification of energy, water, emissions, and e-waste impacts.
Ensure transparency and meaningful participation, including stronger safeguards for AI-generated political content and a formal role for civil society in AI governance.

Read the full submission here: CIPESA Submission on Kenya’s Draft AI and Emerging Technologies Policy.

By Paul Kimumwe |

The past few years have witnessed a growing urgency for frameworks that regulate and harness the development and implementation of new and emerging technologies, especially Generative Artificial Intelligence (Gen AI).

At the international and regional level, the United Nations (UN) and the African Union (AU) have established norms through resolutions, strategies and guidelines to affirm the relationship between technology and human rights, and provide benchmarks for Member States developing rights-respecting AI governance and regulatory frameworks.

In March 2024, the UN adopted a landmark resolution on the promotion of “safe, secure and trustworthy” artificial intelligence (AI) systems that also benefit sustainable development. The resolution also calls upon Member States and other stakeholders “to refrain from or cease the use of artificial intelligence systems that are impossible to operate in compliance with international human rights law or that pose undue risks to the enjoyment of human rights.”

The 2024 resolution reaffirmed that “the same rights that people have offline must also be protected online, including throughout the life cycle of artificial intelligence systems.” It called upon member states to ensure that national AI governance and regulatory frameworks “promote safe, secure and trustworthy artificial intelligence systems” that are inclusive and benefit everyone in an equal manner.

In August 2025, the UN adopted resolution 79/325, establishing the Independent International Scientific Panel on AI and Global Dialogue on AI Governance. It aims to provide a platform to discuss international cooperation, share best practices and lessons learned, and to facilitate open, transparent and inclusive discussions on AI governance. However, a year earlier, in July 2024, the AU adopted the Continental AI Strategy, which emphasises the development of robust governance regimes for AI founded on ethical principles, democratic values, human rights, and the rule of law, in line with the AU Agenda 2063.

Both the UN resolutions on AI and the AU continental strategy came on the backdrop of other AI-related policy guidelines such as Center for AI and Digital Policy’s 2018 Universal Guidelines for AI, the Organization for Economic Cooperation and Development (OECD) 2019 AI Principles / G20 AI Guidelines, the United Nations Education Scientific and Cultural Organization (UNESCO’s) 2021 Recommendation on the Ethics of AI, and the European Union Commission’s (EUC) 2024 European Union AI Act.

Many African countries have been actively developing AI-related laws, policies, and strategies. Rwanda was the first to adopt a national AI policy in 2019, followed by Ghana’s National Artificial Intelligence Strategy in October 2022, Egypt’s National Artificial Intelligence Strategy in January 2025, and Kenya’s own strategy in May 2025. Benin, Côte d’Ivoire, Ethiopia, Mauritius, Nigeria, Tunisia, Zambia, and Zimbabwe are among others that have developed AI policies or strategies. Others, such as Burkina Faso, Guinea, Lesotho, Mali, Namibia, and Uganda, are still at different stages in developing their AI policies or strategies.

A case of history repeating itself?

While all these have been welcome developments in the governance and regulation of AI, studies show that the adoption of international and regional human rights instruments and national laws, policies and strategies is often just the first step in a long process. If not well managed, it often results in provisions that are, although of a progressive nature, are hard to implement and fail to address local needs and realities.

This is because the process of drafting these laws and strategies in many developing contexts is often devoid of meaningful multistakeholder consultations and engagement. Moreover, there has also been a tendency to adopt and replicate models from the global North, whose texts, while progressive, have faced strong resistance from Member States as they sometimes do not align with local contexts and cultural norms.

For example, many African countries, including Algeria, Ethiopia, Cameroon, Kenya, Mauritius, Namibia, Rwanda, South Africa, and Uganda, expressed strong reservations about certain provisions contained in the Protocol to the African Charter on Human and People’s Rights on the Rights of Women in Africa (Maputo Protocol).

Additionally, most of these models are state-centric and grounded in frameworks that create a distinct binary between duty-bearers and rights-holders, but do not articulate how and what each party needs to do to ensure meaningful implementation of the initiatives.

While the state-centric and rights-based approaches may seem attractive, in practice, their relevance in advancing digital rights is often undermined, especially when the prescribed provisions and action points do not align with the country’s current social, economic and political realities. Indeed, cases abound in which initial promises have fizzled over time due to the political leadership’s inaction (and sometimes unwillingness) to fully adopt and implement the resolutions or strategies.

For example, it took almost nine years for the African Union Convention on Cyber Security and Personal Data Protection (Malabo Convention) to enter into force on June 8, 2023, after its adoption in 2014. Indeed, more countries (40) have enacted data protection laws as compared to those that have ratified (16), highlighting a disconnect between national legal reforms and their commitment to continental frameworks. Similarly, the AU Protocol to the African Charter on Human and Peoples’ Rights on the Rights of Persons with Disabilities in Africa, adopted on January 30, 2018, took six years to enter into force, after the 15th ratification was achieved.

More critically, however, the lack of political will often reflect in the absence of clearly defined funding mechanisms for the implementation of these policies and strategies. As a result, even well-designed and progressive frameworks face implementation challenges due to structural flaws and insufficient funding.

For example, while Africa has scored highly in enacting Data Protection laws, which have become central to ongoing AI governance frameworks, one issue affecting their effective implementation is the lack of clear funding mechanisms for the regulatory bodies responsible for oversight and implementation. Other challenges include weak governance structures that deny these oversight bodies financial, decisional and operational independence and place them under the supervision of political appointees rather than parliament.

Designing for Failure?

Apart from Kenya, most African countries that have developed or are in the process of developing an AI strategy or policy do not provide for budgetary allocations or estimates for the implementation of their AI strategies, laws or policies. Countries such as Rwanda provide for a project-level funding framework, while others, such as Egypt and Mauritius, rely on programmatic budgets to fund the implementation of their strategies.

Even then, while implementation of Kenya’s National Artificial Intelligence Strategy (2025–2030) was costed at KSh 152 billion over a period of five years, a review of Kenya’s 2026/27 national budget shows no dedicated funding allocation for the strategy. Instead, the Sh8.6 billion allocated to the ICT sector mainly targets the expansion of broadband access, the strengthening of digital skills, and the digitisation of government services.

Additionally, in countries such as Ethiopia and Rwanda, while the policies provide for the establishment of an implementation body, several functions have been split across different ministries, departments and agencies (MDAs), which, in practice, would pose a significant challenge to meaningful execution.

For example, Rwanda’s AI policy mandates the Responsible AI office under the Ministry of ICT and Innovation to be responsible for effective tech implementation. It also positions the Rwanda Utilities Regulatory Authority (RURA) as the technical regulator responsible for developing ethical AI guidelines and principles, and the National Cyber Security Authority (NCSA) to oversee data protection compliance relevant to AI systems.

In Ethiopia, the policy designates the Ethiopian Artificial Intelligence Institute (EAII) as the national coordinating body responsible for implementation, standards development, and capacity building, and the Ministry of Innovation and Technology is responsible for providing policy oversight. Other sectoral agencies, such as the Ethiopian Communications Authority (ECA), the Ministry of Health, and the National Bank, have mandates over telecommunications and data matters, health-sector-related AI, and financial AI, respectively.

While a multisectoral approach to policy and strategic implementation can improve cohesiveness and legitimacy, the approach is prone to risks such as divergent priorities, internal conflicts, power struggles, and regulatory fragmentation, which are likely to affect how the policies and strategies are executed.

Implications for the Future of AI Governance and Regulation

In many African countries, the development of AI governance and regulatory structures is still in its infancy and presents a unique opportunity for Africans to shape their own destiny on how AI should be developed and deployed in ways that respond to and respect local needs and contexts.

Enactment of AI-specific Laws

In many countries, governments are relying on existing laws, such as data protection, communications, and cyber-related legislation, alongside the AI policies and strategies being developed. Given the evolving nature of AI, countries need to work towards enacting AI-specific laws that clearly define and contextualise AI.

Empowering the Oversight Bodies

As currently structured, many of the existing and proposed oversight bodies are either not yet operational or lack a clear mandate and sufficient resources for effective oversight. Additionally, many of them are situated within fragmented regulatory environments with overlapping responsibilities, which results in uncoordinated implementation. It is important, therefore, that the mandate of the oversight bodies and resources are clearly defined and guaranteed to ensure independence and eliminate the possibility of political interference.

Meaningful Stakeholder Participation

Having empowered stakeholders who are meaningfully engaged and participate in the development processes for policies, laws and strategies is critical to ensuring that the resulting instruments address real needs, are people-centred and implementable, and have government buy-in, as reflected in the government’s funded priorities.

Adopting a Human Rights-Centred Approach

A 2025 study by CIPESA shows that in many countries, the adoption of a human rights-centred approach to AI governance remains aspirational due to gaps in implementation, technical capacity, and stakeholder engagement in policy development and implementation. It is important, therefore, that current efforts prioritise safeguarding fundamental human rights and freedoms, enhancing human capabilities over replacement, and ensuring meaningful human control, transparency, fairness, and inclusivity in AI systems.

Tag Archives: AI Policy

Rethinking Africa’s Approach to the Politics of AI Governance and Regulation