Preamble

The Africa AI Charter is the Rosetta Stone for African AI governance. This means that like the Rosetta Stone, which enabled the deciphering of ancient languages, the Charter would serve as the foundational framework or interpretive key through which African states, regulators, businesses, and citizens understand, coordinate, and implement AI governance across the continent.

The shared definitions and standards could unlock cooperation among countries with different legal systems and levels of technological development. It will demonstrably provide the crucial key for deciphering, understanding, governing, or advancing artificial intelligence across nations in Africa.

Therefore, we, the signatories to this Charter governments, institutions, enterprises, researchers, innovators, and citizens of Africa and its diaspora.

RECOGNIZING that artificial intelligence has become economic infrastructure, security infrastructure, and geopolitical infrastructure, and that the terms on which it is created and deployed will shape the prosperity, dignity, and sovereignty of African peoples for generations;

AFFIRMING that Africa must not be a passive consumer of intelligence systems designed elsewhere, trained on data extracted from its people, and governed by rules written without its participation, but must be an architect of the technologies that will define this century;

GROUNDED in the philosophy of Ubuntu, I am because we are, which holds that the worth of any technology is measured by its contribution to the flourishing of the community, the dignity of the person, and the continuity of culture;

RECALLING the Constitutive Act of the African Union, the African Charter on Human and Peoples' Rights, the African Union Convention on Cyber Security and Personal Data Protection (Malabo Convention), the African Union Continental AI Strategy, Agenda 2063, and the Agreement Establishing the African Continental Free Trade Area;

CONSCIOUS that risk is not a property of a technology in isolation but a relationship between a technology and the institutional environment in which it operates, and that African AI governance must therefore begin with honest assessment of African institutional contexts rather than the importation of risk categories calibrated elsewhere;

CONVINCED that the data generated by African populations constitutes a continental strategic asset; that African languages, knowledge systems, and cultural heritage must be represented within the world's intelligence systems; and that trust, verifiable, institutional, and earned, is the true foundation of any digital economy;

MINDFUL that African civilisations encoded computable procedures in story, proverb, and ceremony long before the modern era, a heritage of alagorithms, ancestral rules that compute, and that this heritage belongs among the foundational inputs to the intelligence systems of this century;

DETERMINED to build the research capacity, compute infrastructure, technical education, and institutional architecture that make African AI creation possible, so that the continent may govern what it understands and build what it governs;

HAVE AGREED to adopt this Charter as the normative framework governing the creation, deployment, and implementation of artificial intelligence in Africa, and to be guided by its principles in law, in policy, in enterprise, and in practice.

The Alagorithm

/ə-LAG-ə-rith-əm/ · noun · allegory + algorithm — an ancestral rule that computes.

Long before the world borrowed the word algorithm from the mathematician al-Khwarizmi, African civilizations were encoding computable procedures in story, proverb, and ceremony. A rule did not need a manual to survive; it was carried in allegory — performed, danced, recited — and it executed, generation after generation, with the reliability of code. The Charter names this inheritance the alagorithm: ancestral wisdom expressed as procedure, procedure preserved as story.

Consider the counterclockwise governance dance at the Fon's Palace in Bamenda, Cameroon: a ceremony that is also an instruction set — ordering movement, precedence, and authority in steps any participant can execute and any generation can inherit. Procedures outlive their explanations. That is the alagorithm at work.

The Charter gives this concept the force of an instrument: the alagorithm is a defined term of Article 2, its heritage is invoked in the Preamble, and Article 11(5) commits States Parties to document, protect, and represent Africa's alagorithms within AI systems — under the same consent, attribution, and benefit-sharing protections that guard all indigenous knowledge.

The term alagorithm was coined by Dr. Stanley Fongod, Founding Convener. Canonical definition at alagorithm.org.

Purpose and Scope

1. This Charter establishes the principles, obligations, rights, and institutional arrangements governing the creation, development, deployment, procurement, and implementation of artificial intelligence systems within Africa, and by African entities operating abroad.
2. This Charter applies to States, public bodies, private enterprises, research institutions, civil society organizations, and individual developers and deployers of AI systems, to the extent of their respective roles in the AI lifecycle.
3. The Charter is designed to operate in harmony with national constitutions, the instruments of the African Union, and international law, and to serve as a model framework for national AI legislation and regulation across the continent.
4. Nothing in this Charter shall be interpreted to restrict rights and protections already guaranteed under national or international law; where standards conflict, the standard most protective of human dignity shall prevail.

Definitions

For the purposes of this Charter:

Sovereignty and Strategic Autonomy

1. Artificial intelligence is hereby recognized as economic infrastructure, national security infrastructure, and strategic geopolitical infrastructure of the African continent.
2. States Parties shall pursue sovereign AI capability, including national and regional compute infrastructure, sovereign datasets, indigenous research capacity, and domestic technical talent, as a matter of strategic policy.
3. Dependency on foreign AI systems for critical public functions, including identity, payments, health records, judicial administration, and electoral processes, shall be assessed, disclosed, and progressively reduced through capability development and diversification.
4. No agreement with a foreign state, corporation, or multilateral body shall cede permanent control over African critical data infrastructure, nor grant exclusive extraterritorial jurisdiction over data generated within African territories, except under terms ratified through the constitutional processes of the State concerned.

Human Dignity and the Ubuntu Ethical Standard

1. The human person is the measure of all intelligent systems. No consideration of efficiency, profit, or State interest shall justify the creation or deployment of an AI system that undermines human dignity.
2. All AI creation and implementation under this Charter shall be evaluated against the Ubuntu Ethical Standard: whether the system strengthens the community, honours the individual, preserves the inherited wealth of culture and language, and safeguards the interests of future generations.
3. AI systems shall be designed and deployed to augment human judgement, capability, and agency, and not to displace human moral responsibility. Legal and moral accountability for the consequences of an AI system shall at all times rest with identifiable natural or legal persons.

Non-Discrimination, Inclusion, and Linguistic Justice

Particular attention shall be paid to preventing discrimination arising from proxy variables, historical inequalities, and data deficiencies that disproportionately affect African populations. This is an important issue in algorithmic fairness literature.

1. AI systems created or implemented in Africa shall not discriminate, in purpose or in effect, on the basis of race, ethnicity, tribe, sex, gender, religion, language, disability, age, socio-economic status, or political opinion.
2. Developers shall identify, assess, and mitigate bias arising from unrepresentative training data, including the systematic underrepresentation of African populations, dialects, names, geographies, and social contexts within global datasets.
3. Recognising that African languages are repositories of knowledge, identity, and reasoning, States Parties and developers shall invest in the creation of high-quality datasets, benchmarks, and models for African languages. Public-facing AI services shall progressively provide access in the principal languages of the populations they serve.
4. Rural communities, women, young people, persons with disabilities, and workers in the informal economy shall be deliberately included in AI access initiatives, AI literacy programmes, and the economic benefits arising from AI implementation.

Data as a Continental Strategic Asset

1. Data generated by African persons, enterprises, ecosystems, and territories constitutes a strategic asset of the continent and of the peoples from whom it originates.
2. The extraction, transfer, or exploitation of African data for the training or development of AI systems without a lawful basis, transparency, and equitable value-sharing arrangements is contrary to the principles of this Charter.
3. States Parties shall establish national data governance frameworks consistent with the African Union Convention on Cyber Security and Personal Data Protection (Malabo Convention) and this Charter. Such frameworks shall include: (a) registers of significant datasets; (b) data-sharing agreements subject to public oversight and accountability; and (c) mechanisms for the fair and equitable sharing of benefits arising from the commercial exploitation of population-scale data.

Cross-Border Data Flows

1. Cross-border data flows within Africa shall be facilitated under harmonised continental standards, in support of the digital protocols of the African Continental Free Trade Area, so that African data may contribute to the development of African AI.
2. The transfer of sovereign datasets beyond the continent shall require: (a) a lawful basis; (b) documented technical and organisational safeguards; (c) enforceable accountability obligations on the recipient; and (d) where the datasets have been designated as critical, the prior approval of the competent national authority.
3. Data localisation requirements shall be proportionate, justified by legitimate security or public-interest considerations, and designed to strengthen domestic capability rather than to shield poor practices from scrutiny or accountability.

Sovereign Datasets and National Data Estates

1. Each State Party shall identify, classify, and designate its sovereign datasets, which may include, inter alia, population health data, agricultural and climatic data, geospatial information, linguistic corpora, biometric and identity data, and financial-system data.
2. Sovereign datasets shall be governed by documented stewardship arrangements specifying: (a) custodianship and accountability; (b) conditions of access and use; (c) security and integrity standards; (d) permitted uses in AI training and development; and (e) revenue-sharing or other benefit-sharing arrangements.
3. States Parties are encouraged to establish national and regional data trusts, data commons, and open-data programmes that make non-sensitive public data available to African researchers, innovators, and enterprises on fair, transparent, and non-discriminatory terms, while prioritising African access and value creation.

Privacy and Personal Data Protection

1. Persons shall have the right, subject to national law, to access, correct, delete, and where appropriate port their personal data processed by AI systems.

This aligns with the:

GDPR;

Malabo Convention;

Convention 108+.

Every person has the right to the protection of personal data concerning them. The creation and implementation of AI systems shall comply with applicable data protection laws and, in the absence of such laws, shall adhere, at a minimum, to the standards established by the Malabo Convention.

2. Personal data used in AI training or operation shall: (a) be collected and processed lawfully and fairly; (b) be limited to what is necessary and proportionate for the stated purpose; (c) be protected against unauthorised access, misuse, and re-identification; and (d) where feasible, be processed using privacy-preserving techniques, including anonymisation, pseudonymisation, and other appropriate safeguards.
3. The use of AI systems for the biometric identification of persons in publicly accessible spaces shall be prohibited except where expressly authorised by law and subject to: (a) prior judicial or independent authorisation; (b) a demonstrated necessity and proportionality assessment; and (c) appropriate safeguards against abuse, discrimination, and unlawful surveillance.

Responsible Development Obligations

1. Developers of frontier AI models shall conduct cybersecurity testing, misuse assessments, and evaluations of systemic risk, including risks of large-scale misinformation, cyber abuse, and autonomous replication where relevant. This reflects current international discussions around frontier AI. This reflects current international discussions around frontier AI.
2. Developers shall implement risk management measures throughout the entire AI lifecycle. Such measures shall include:

(a) documented design objectives and intended uses;

(b) records of data provenance and data governance practices;

(c) assessment of reasonably foreseeable misuse and unintended consequences;

(d) evaluation of impacts on safety, security, fundamental rights, and the environment; and

(e) pre-release testing and validation proportionate to the system's capabilities, scale, and potential risks.

3. Developers of general-purpose AI models and frontier AI models made available within Africa shall publish sufficient technical documentation to enable deployers, regulators, and independent auditors to understand:

(a) the model's capabilities and limitations;

(b) the categories and sources of training data, to the extent consistent with legitimate confidentiality and intellectual property interests;

(c) known failure modes and material risks; and

(d) the safeguards and mitigation measures implemented by the developer.

4. Developers shall establish accessible mechanisms for incident reporting and vulnerability disclosure and shall notify the competent national authority, without undue delay, of serious incidents arising from their systems, including incidents that result in, or create a significant risk of:

(a) harm to health or safety;

(b) violations of fundamental rights;

(c) significant economic or environmental harm; or

(d) disruption to critical infrastructure or essential public services.

5. Developers of frontier AI models whose capabilities may pose systemic risks shall conduct appropriate safety evaluations and implement proportionate safeguards, including external testing and independent review where required by national law.

African Language, Knowledge, and Cultural Representation

1. The creation of AI in Africa shall recognise African languages, oral traditions, indigenous knowledge systems, and cultural heritage as foundational inputs and protected assets rather than residual or marginal categories.
2. Indigenous and community knowledge incorporated into AI systems shall be used only with the free, prior, and informed consent of the communities that hold such knowledge, on terms agreed with those communities and, where commercial value is derived, with appropriate attribution and equitable benefit-sharing arrangements.
3. States Parties shall support the development of African language corpora, speech datasets, translation resources, and evaluation benchmarks as public digital infrastructure and shall protect such assets against uncompensated extraction, misappropriation, and misuse.
4. States Parties shall promote the preservation and digitisation of endangered African languages and shall encourage the development of AI systems capable of supporting linguistic diversity and cultural continuity across the continent.
5. States Parties shall recognise Africa’s alagorithms, the computable rules and procedures preserved in proverb, ceremony, and communal practice, as part of the continent’s intelligence heritage, to be documented, protected, and represented within AI systems in accordance with the consent, attribution, and benefit-sharing requirements of this Article.

Research Capacity and Talent Development

1. States Parties shall invest in African AI research capacity, including university research programmes, centers of excellence, national research institutes, doctoral and postdoctoral training, and continental research networks, with the objective that Africa contributes to, and not merely consumes, the global scientific frontier.
2. AI literacy and digital skills shall be progressively integrated into national education systems, technical and vocational education and training, and public-sector professional development programmes.
3. States Parties and institutions shall create conditions conducive to the retention, circulation, and return of African AI talent, including:

(a) sustainable research funding;

(b) competitive research and innovation ecosystems;

(c) diaspora engagement frameworks;

(d) mobility and exchange programmes; and

(e) recognition and support for African-based research careers.

4. States Parties are encouraged to promote equitable participation in AI research and education, with particular attention to women, young people, persons with disabilities, and historically underrepresented communities.

Compute, Energy, and Infrastructure

1. States Parties shall promote participation in the semiconductor, advanced computing, and digital infrastructure value chains in order to reduce strategic dependencies and enhance technological resilience.
2. Access to computational infrastructure is a prerequisite for AI sovereignty. States Parties shall pursue, individually and through regional cooperation, the development of African computational capacity, including national and shared regional data centers, high-performance computing facilities, cloud infrastructure, and secure digital connectivity.
3. The development of AI infrastructure shall be integrated with energy policy and long-term sustainability planning, prioritising reliable, affordable, and increasingly renewable sources of power, while taking due account of water consumption, land rights, environmental impacts, and climate resilience.
4. States Parties shall encourage the establishment of public-interest compute programmes and shall provide equitable access to computational resources for African universities, research institutions, start-ups, and public-interest research initiatives.
5. . Regional cooperation in computational infrastructure shall be encouraged in order to reduce costs, enhance resilience, and promote the equitable distribution of technological capacity across the continent.

Intellectual Property, Openness, and Fair Innovation

1. This Charter affirms a balanced innovation framework that:

(a) protects legitimate intellectual property rights;

(b) supports open-source and open-science AI development;

(c) promotes knowledge sharing and collaborative innovation; and

(d) safeguards the rights and interests of African creators whose works are used in AI development and training.

2. The use of African creative, journalistic, scholarly, and cultural works in the training of AI models shall respect applicable intellectual property and related rights. States Parties shall develop licensing, remuneration, and collective rights management frameworks appropriate to their creative and knowledge economies.
3. Publicly funded AI research, datasets, and models shall, by default and subject to legitimate security, privacy, and commercial confidentiality considerations, be released under open licences that permit African researchers, educational institutions, and enterprises to access, use, adapt, and build upon them.
4. Nothing in this Charter shall be interpreted as preventing States Parties from adopting measures necessary to protect traditional knowledge, cultural expressions, or other forms of intellectual heritage that require special protection under national or international law.

Risk-Based Classification

1. AI systems implemented in Africa shall be classified according to their level of risk, namely: (a) Prohibited Practices; (b) High-Risk Systems; (c) Limited-Risk Systems; and (d) Minimal-Risk Systems, as further elaborated in Annex A.
2. The following practices shall be prohibited: (a) AI systems used by public authorities for social scoring that results in unjustified or disproportionate detriment to individuals or groups; (b) AI systems that exploit the vulnerabilities of children, older persons, persons with disabilities, or persons in situations of distress in a manner likely to cause physical, psychological, or economic harm; (c) AI systems that employ deceptive or manipulative techniques that materially distort behaviour and are likely to cause significant harm; and (d) indiscriminate or untargeted mass surveillance that is inconsistent with Article 9(3) and fundamental rights protections.
3. Risk classification shall take into account the African institutional context in which a system operates, including the effectiveness of oversight bodies, the availability of redress mechanisms, levels of digital literacy, and the resilience of affected institutions, in accordance with the principle of context-sensitive risk assessment.

Obligations for High-Risk Systems

1. Prior to the deployment of a High-Risk AI System, the deployer shall conduct and document an AI Impact Assessment covering:

(a) the intended purpose and scope of the system;

(b) affected individuals and communities;

(c) potential harms and adverse impacts;

(d) bias and discrimination risks;

(e) data quality and governance;

(f) cybersecurity and resilience; and

(g) mitigation and monitoring measures.

2. High-Risk AI Systems shall:

(a) be registered with the competent national authority and reassessed following any material modification;

(b) be subject to meaningful human oversight, including the authority to override, suspend, or discontinue the system;

(c) maintain logs and records sufficient to reconstruct significant decisions and incidents; and

(d) comply with retention periods prescribed by national law.

3. High-Risk AI Systems shall meet documented standards of accuracy, robustness, reliability, and cybersecurity appropriate to their intended context of use and shall be validated before deployment and continuously monitored thereafter.
4. Where a High-Risk AI System is procured from a third party, the procuring entity shall remain responsible for ensuring compliance with the requirements of this Article.

AI in the Public Sector

1. Public authorities shall not rely solely upon proprietary claims of confidentiality to avoid disclosure necessary for accountability, judicial review, or independent oversight.

This is increasingly important because governments frequently procure "black box" AI.

Public bodies deploying AI systems shall do so pursuant to published policies and shall maintain a publicly accessible register of AI systems used in consequential decision-making.

2. No person shall be subjected to a solely automated decision by a public authority that produces legal effects or similarly significant consequences without meaningful human review and the right to contest the decision.
3. AI shall not diminish or replace the constitutional accountability of public officials. The deployment of AI in policing, justice, immigration, social protection, taxation, and other essential public services shall be subject to heightened scrutiny, independent oversight, and periodic public audit.
4. Electoral processes shall be protected against AI-enabled manipulation, including:

(a) undisclosed synthetic media intended to mislead voters;

(b) coordinated inauthentic amplification and disinformation campaigns;

(c) automated voter suppression systems; and

(d) other forms of digital interference that undermine electoral integrity.

5. States Parties shall adopt measures to safeguard democratic processes from foreign and domestic AI-enabled information operations that threaten constitutional order or public trust.

Obligations of Private Enterprises

1. Enterprises whose activities involve the large-scale processing of personal data or the deployment of High-Risk AI Systems should designate an individual or body responsible for AI governance and compliance proportionate to their size, capabilities, and risk profile, including:

(a) clear allocation of responsibility for AI risks at senior management level;

(b) staff training and capacity-building;

(c) supplier and third-party due diligence; and

(d) internal procedures for risk management and incident reporting.

2. Enterprises shall be transparent when individuals are interacting with an AI system rather than a human and when content has been synthetically generated or materially altered by AI in circumstances where such information is relevant to informed decision-making.
3. Multinational providers offering AI systems in African markets shall comply with standards of safety, transparency, redress, and accountability that are no less protective than those applied in their principal markets and shall designate a legal representative in the jurisdictions in which they operate.
4. Enterprises whose AI systems pose systemic or significant societal risks may be required by national law to undergo independent audits or conformity assessments.

Public Procurement of AI

1. The public procurement of AI systems shall require:

(a) disclosure of training data categories and known limitations;

(b) local testing and validation against the populations and environments to be served;

(c) contractual rights of audit and inspection;

(d) exit, interoperability, and data portability provisions;

(e) security vetting for systems performing critical functions; and

(f) clear allocation of liability and maintenance obligations.

2. Procurement frameworks shall, consistently with applicable trade obligations, encourage:

(a)African-built systems and services;

(b) African data hosting and processing;

(c) local skills and technology transfer; and

(d) joint ventures and partnerships that strengthen domestic capabilities and regional value chains.

3. No public authority shall procure an AI system whose provider refuses independent evaluation of claims material to the contract or declines to provide information necessary for regulatory oversight.

Rights of Persons Affected by AI

To be protected against unlawful profiling and manipulative practices that materially impair autonomy or exploit vulnerability. Every person within the scope of this Charter shall have the right:

(a) to know when they are subject to a consequential AI-assisted decision;

(b) to receive a meaningful explanation of such a decision in clear and accessible language;

(c) to obtain human review of, and to contest, such a decision;

(d) to the protection of their personal data, identity, image, and voice, including protection against non-consensual synthetic reproduction or impersonation;

(e) to effective redress and remedy for harm caused by an AI system, as provided in Article 23; and

(f) to freedom from unlawful discrimination arising from the use of AI systems.

Transparency and Explainability

1. Transparency obligations shall be proportionate to risk: the greater the potential impact of a system on individuals, communities, or public interests, the greater the duty of disclosure and explanation.
2. Explanations provided to affected persons shall be truthful, intelligible to a non-specialist, and sufficient to enable effective contestation and review. The technical opacity of a model shall not excuse institutional opacity in decision-making.
3. Synthetic media generated or substantially modified by AI shall be clearly disclosed in news reporting, political communications, advertising, and official communications, except where such disclosure would undermine legitimate law-enforcement activities authorised by law.

Human Oversight and the Limits of Automation

1. The decision to take human life shall never be delegated to a fully autonomous weapon system operating without meaningful human control. This reflects the growing international consensus on autonomous weapons. Meaningful human oversight shall be mandatory for High-Risk AI Systems and for all consequential decisions taken in public sector.
2. Oversight arrangements shall guard against automation bias. Nominal human involvement, without the competence, information, time, or authority necessary to intervene, shall not satisfy the requirements of this Charter.
3. Certain decisions shall remain exclusively human and shall not be delegated to autonomous systems, including:

(a) the imposition of criminal penalties;

(b) decisions concerning asylum, refugee status, deportation, or extradition;

(c) the authorisation of lethal force; and

(d) any other decisions designated by national law as requiring inherently human judgement.

Redress, Remedy, and Liability

1. States Parties shall ensure accessible, affordable, and timely avenues of complaint and redress for persons harmed by AI systems, through regulators, ombuds institutions, and the courts.
2. Liability regimes shall allocate responsibility across the AI value chain, including developers, providers, deployers, operators, and, where appropriate, procuring entities, so that no injured person is left without an effective remedy by reason of technical complexity or fragmented accountability.
3. Where an AI system has caused harm and information asymmetry prevents the injured person from proving fault, national law should provide for:

(a) appropriate disclosure obligations;

(b) powers for courts or regulators to compel access to relevant information; and

(c) for High-Risk Systems, rebuttable presumptions in favour of the injured person where justice so requires.

Labour, Livelihoods, and the Future of Work

1. States Parties should monitor and mitigate labour-market concentration, technological displacement, and inequalities arising from the uneven distribution of AI benefits.
2. The implementation of AI in the workplace shall respect labour rights, including consultation with workers or their representatives before the deployment of systems that materially affect recruitment, evaluation, promotion, remuneration, discipline, or dismissal.
3. Algorithmic management of workers, including platform and gig workers, shall be transparent, explainable, contestable, and consistent with principles of decent work and fair treatment.
4. States Parties shall prepare for AI-driven labour transitions through reskilling and lifelong learning programmes, adaptation of social protection systems, and the deliberate cultivation of AI-enabled industries in which Africa's young and growing workforce possesses comparative advantages.

Health

1. Health AI systems shall be subject to post-market monitoring and mechanisms for reporting adverse events and clinical incidents.
2. AI systems used in diagnosis, triage, treatment recommendations, public health decision-making, or healthcare resource allocation shall be classified as High-Risk AI Systems and shall be clinically validated on populations that are representative of those they are intended to serve.
3. Health AI shall support, and not replace, the professional judgement and responsibility of qualified healthcare practitioners and shall be deployed with due regard to the realities of African health systems, including connectivity constraints, workforce shortages, and referral capacity.
4. Health data constitutes a category of highly sensitive sovereign data. Its use in AI development shall comply with Part II of this Charter and shall be subject to enhanced privacy, consent, and security safeguards.

Education

1. AI in education shall serve pedagogical objectives by expanding access, supporting educators, and personalising learning, without narrowing curricula, entrenching surveillance of learners, or undermining the development of human judgement and critical thinking.
2. AI literacy, including the critical evaluation of AI outputs and the responsible use of AI tools, shall be treated as a core competency at all levels of education.
3. Educational data relating to minors shall receive the highest level of protection, and its commercial exploitation for unrelated purposes shall be prohibited.

Agriculture, Finance, and Economic Inclusion

States Parties shall encourage the development of AI applications that support financial inclusion, informal-sector productivity, and small and medium-sized enterprises. This is particularly important for African economies.

1. AI in agriculture shall be developed with and for African farmers, respecting land and agricultural data rights, supporting smallholder productivity and climate adaptation, and avoiding dependence on proprietary ecosystems that undermine farmer autonomy.
2. AI-driven credit scoring, insurance pricing, and financial services shall be explainable, contestable, and regularly monitored for discriminatory effects. Such systems shall be designed to expand, rather than restrict, financial inclusion, recognising that trust and accessibility are fundamental to Africa's financial deepening.
3. States Parties shall encourage AI applications addressing continental priorities: food security, energy access, logistics, healthcare delivery, public administration, and climate resilience.
4. Corrected legal terminology and improved consistency with international instruments such as the EU AI Act, OECD AI Principles, and the Council of Europe Framework Convention on AI.
5. Added protection against deepfake impersonation and AI-enabled electoral interference.
6. Expanded labour protections to include recruitment and promotion decisions, not merely dismissal.
7. Clarified that educational data on minors cannot be commercially exploited for unrelated purposes, avoiding unintended restrictions on legitimate educational services.
8. Recognised healthcare resource allocation systems as High-Risk AI Systems, reflecting current international regulatory practice.
9. Strengthened provisions on algorithmic accountability, systemic risk, and access to effective remedies.

National AI Authorities

National AI authorities should possess multidisciplinary expertise, including law, computer science, ethics, economics, social science, and human rights.

1. Each State Party shall designate or establish a competent national authority for artificial intelligence that is adequately resourced, technically capable, and operationally independent. The authority shall be responsible for, inter alia:

(a) the registration and oversight of High-Risk AI Systems;

(b) the development and enforcement of standards and guidance;

(c) market surveillance and regulatory supervision;

(d) incident reporting and response; and

(e) public information and awareness.

2. National AI authorities shall cooperate and coordinate with data protection authorities, sectoral regulators, competition authorities, consumer protection bodies, and cybersecurity agencies in order to avoid regulatory fragmentation, duplication, and gaps in oversight.
3. States Parties shall adopt and periodically update national AI strategies aligned with this Charter and shall report regularly on their implementation, including progress, challenges, and emerging risks.
4. States Parties are encouraged to establish mechanisms for public consultation and stakeholder participation in the formulation and implementation of national AI policies.

Continental Coordination

1. States Parties shall work towards the development of interoperable African standards and, where appropriate, common certification mechanisms and regulatory sandboxes.
2. States Parties shall pursue the harmonisation of AI governance through the African Union and the Regional Economic Communities, including:

(a) mutual recognition of conformity assessments and certifications;

(b) the establishment of shared incident and vulnerability databases;

(c) cooperation in research, education, and computational infrastructure;

(d) the development of common standards and technical guidance; and

(e) coordination of regulatory approaches and capacity-building initiatives.

3. Africa shall participate in global AI governance through coordinated and well-prepared positions supported by technical expertise, so that the rules governing the age of artificial intelligence are not shaped in the continent's absence.
4. States Parties shall encourage South-South cooperation and strategic partnerships that advance African technological sovereignty, scientific capacity, and equitable participation in the global digital economy.

The Charter Secretariat and the Africa AI Charter & Center

1. The Africa AI Charter & Centre shall serve as the Secretariat of this Charter and shall perform the following functions:

(a) custodianship of the Charter and its interpretative guidance;

(b) independent research and publication of the Africa AI Readiness Index;

(c) policy advice and technical assistance to governments and institutions;

(d) capacity development through the Africa AI Academy and related programmes; and

(e) the convening of the annual Africa AI Summit and the network of signatories.

2. The Secretariat shall:

(a) maintain a public register of signatories;

(b) publish an annual State of AI Governance in Africa report;

(c) facilitate peer review and the exchange of best practices among States Parties; and

(d) support the implementation and periodic review of this Charter.

3. The Secretariat shall operate in a transparent and accountable manner with respect to its governance, funding, and activities. Its research independence and operational integrity shall be protected from undue political, donor, or commercial influence.
4. Nothing in this Charter shall prevent the African Union or States Parties from establishing additional institutions or mechanisms necessary to advance the objectives of this Charter.

Compliance, Audit, and Conformity Assessment

Independent auditors shall be free from conflicts of interest and shall possess the technical competence necessary to assess the systems under review.

1. High-Risk AI Systems shall undergo an appropriate conformity assessment prior to deployment and shall be subject to periodic audits thereafter. Such assessments may be conducted by the deployer, accredited third parties, or the competent national authority, depending on the level of risk and the nature of the system.
2. States Parties shall promote and accredit African capacity in AI auditing, testing, assurance, and conformity assessment, so that regulatory compliance contributes to the development of domestic expertise rather than creating new forms of technological dependency.
3. Signatory enterprises and institutions shall periodically certify their adherence to this Charter in accordance with procedures established by the Secretariat, and material breaches shall be disclosed in a transparent manner.
4. States Parties shall encourage the development of standards, benchmarks, and certification schemes that are interoperable with international frameworks while reflecting African priorities and circumstances.

Sanctions, Suspension, and Safeguard Measures

In determining sanctions, authorities shall consider the gravity of the violation, the degree of negligence or intent, previous infringements, cooperation with regulators, and the measures taken to mitigate harm.

1. National law shall provide for effective, proportionate, and dissuasive sanctions for violations of obligations arising under this Charter, including:

(a) administrative fines;

(b) corrective orders and mandatory remediation;

(c) suspension or restriction of deployment;

(d) withdrawal of systems from the market; and

(e) any other measures necessary to protect the public interest.

2. The competent national authority may order the immediate suspension or restriction of any AI system that presents a serious and imminent risk to health, safety, fundamental rights, democratic processes, or critical infrastructure, pending further investigation and remedial action.
3. The Secretariat may suspend or withdraw the status of a signatory entity that persistently and materially breaches this Charter, following due notice and an opportunity to respond, and shall publish the reasons for such determinations.
4. Sanctions and enforcement measures shall respect the principles of due process, proportionality, transparency, and the right to an effective remedy.

Adoption, Accession, and Amendment

1. This Charter shall be open for signature or accession by States, public institutions, enterprises, universities, research institutions, civil society organisations, and individuals, each assuming obligations appropriate to its role and capacity.
2. Amendments to this Charter may be proposed by any signatory and shall be considered at a Review Conference convened by the Secretariat. The Review Conference shall meet at least once every three years, or more frequently where necessary, to ensure that the Charter remains responsive to technological, social, and geopolitical developments.
3. States Parties are encouraged to domesticate and implement the principles of this Charter through national legislation, regulation, standards, public procurement policies, and institutional reforms.
4. Amendments adopted by the Review Conference shall take effect in accordance with procedures established by the Conference and the implementing arrangements of the Charter.

Interpretation, Language, and Entry into Force

1. This Charter shall be interpreted in good faith, in accordance with its ordinary meaning, in the light of its Preamble, the Ubuntu Ethical Standard, and its overarching objective of promoting African agency, dignity, and sovereignty in the age of artificial intelligence.
2. The Charter shall be published in the working languages of the African Union. Translations into African languages shall be actively promoted in order to broaden accessibility and participation. All authentic language versions adopted by the Secretariat shall be equally authoritative.
3. This Charter shall enter into force for each signatory upon signature, accession, or formal adoption, in accordance with its internal procedures.
4. This Charter is intended to function as a living and evolving framework, whose authority and influence shall grow through its implementation, adoption, and continuous development by African institutions, States, enterprises, and citizens.

Risk Classification Schedule

Systemic-Risk General-Purpose AI

Frontier models whose capabilities or scale may create widespread economic, social, security, or democratic risks. This category now appears in the EU AI Act and would future-proof the Charter.

This Annex elaborates the risk classification framework established under Article 15. The classification of an AI system shall take into account its intended purpose, reasonably foreseeable uses, scale of deployment, potential impact, and the national institutional context in which it operates, in accordance with the principle of context-sensitive risk assessment.

Dynamic Classification

The classification of an AI system shall not be determined solely by its technical characteristics but also by its intended purpose, context of deployment, scale, and potential consequences.

An AI system classified as Minimal-Risk or Limited-Risk may be reclassified as a High-Risk System where its deployment creates significant risks to health, safety, fundamental rights, democratic processes, or critical infrastructure.

National AI authorities may issue guidance, designate additional categories of High-Risk Systems, or impose supplementary safeguards in response to technological developments or emerging risks.

The Secretariat shall periodically review this Annex and may recommend amendments to ensure that the risk classification framework remains responsive to scientific advances, evolving uses of AI, and the particular needs and circumstances of African societies.

Implementation Benchmarks for States Parties

To promote accountability, measure progress, and facilitate peer learning, States Parties shall report periodically on their implementation of this Charter against the benchmarks set out below. These benchmarks shall inform the Africa AI Readiness Index maintained by the Secretariat and may be supplemented by additional indicators adopted from time to time.

1. Governance and Regulatory Capacity: States Parties should demonstrate:

(a) the publication and periodic updating of a national AI strategy aligned with this Charter;

(b) the designation or establishment of a competent national AI authority that is adequately resourced, technically capable, and operationally independent;

(c) the establishment and maintenance of a register of High-Risk AI Systems;

(d) the adoption of national standards, guidance, and procedures for AI governance and oversight; and

(e) evidence of coordination among AI, data protection, cybersecurity, competition, and sectoral regulatory authorities.

2. Data Governance and Sovereignty:

States Parties should demonstrate:

(a) the enactment and effective implementation of data protection legislation consistent with the principles of this Charter and the Malabo Convention;

(b) the identification and designation of sovereign datasets, together with documented stewardship and governance arrangements;

(c) the adoption of lawful and secure frameworks governing cross-border data transfers;

(d) the establishment of data-sharing mechanisms and, where appropriate, benefit-sharing arrangements; and

(e) measurable improvements in data quality, accessibility, and interoperability.

3. Research, Infrastructure, and Capability Development:

States Parties should demonstrate:

(a) the establishment of national or regional programmes providing access to computational infrastructure;

(b) the integration of AI literacy and digital skills into tertiary, technical, and vocational education;

(c) measurable growth in domestic AI research, innovation, and enterprise formation;

(d) investment in research institutions, centres of excellence, and talent development programmes; and

(e) initiatives to attract, retain, and engage African scientific and technical talent, including members of the diaspora.

4. Rights Accountability, and Redress:

States Parties should demonstrate:

(a) the existence of accessible and effective complaint and redress mechanisms for persons affected by AI systems;

(b) the publication and maintenance of public-sector AI registers;

(c) evidence of enforcement actions and remedies for AI-related harms;

(d) measures to promote transparency, public awareness, and digital literacy; and

(e) the availability of judicial or administrative remedies for violations of rights protected under this Charter.

5. Participation and International Engagement:

States Parties should demonstrate:

(a) active participation in African Union and Regional Economic Community initiatives relating to artificial intelligence;

(b) meaningful engagement in global AI governance forums and standard-setting processes;

(c) the development of informed and coordinated national positions on international AI governance issues;

(d) support for regional cooperation, knowledge sharing, and cross-border research initiatives; and

(e) contributions to the development of a common African voice in global digital governance.

5. Public Awareness and Inclusion:

States Parties should demonstrate:

national AI literacy programmes;

initiatives promoting participation by women and underrepresented groups;

public consultations on AI governance;

publication of plain-language guidance on AI rights;

support for civil society engagement in AI governance

Review of Benchmarks

The Secretariat shall periodically review these benchmarks and may recommend additional indicators to reflect technological developments, emerging risks, and evolving continental priorities.

States Parties are encouraged to publish their progress reports and implementation data in order to promote transparency, accountability, and the exchange of best practices.

Adoption and Signature

IN WITNESS WHEREOF, the undersigned hereby adopt this Charter and solemnly undertake to be guided by its principles in the creation, development, deployment, and governance of artificial intelligence in Africa.

Done at ______________________________

On this ______ day of ____________________, ______.

For States Parties

Name of State: ______________________________________

Name and Title of Authorised Representative:

Signature: ___________________________________________

Official Seal: ________________________________________

For Institutions and Organisations

Name of Institution or Organisation:

Name and Title of Authorised Representative:

Signature: ___________________________________________

Official Seal (where applicable): _______________________

For Individual Signatories

Name: _______________________________________________

Nationality: __________________________________________

Organisation (if any): _________________________________

Signature: ___________________________________________

Date: ________________________________________________

Signatory

Name, Institution, and Capacity