Steven W Kayuni
Office of President and Cabinet, Malawi Government, Lilongwe
skayuni@lawyer.com

Introduction

Artificial intelligence (AI) and advanced surveillance technologies are increasingly deployed in armed conflicts across Africa. Facial recognition, biometric registration, communications interception, predictive analytics and drone surveillance are now commonly used in counter-terrorism and internal security operations. These developments raise complex questions for international humanitarian law (IHL) and international criminal law (ICL), particularly in relation to attribution of responsibility, evidentiary standards and civilian protection. The African Union (AU) has taken steps to develop continental norms on digital governance.¹ However, it lacks a conflict-specific regulatory framework addressing the use of AI-enabled surveillance during armed conflict. This article examines the implications of that gap and outlines practical regulatory options relevant to war crimes practitioners.

AI-enabled surveillance and armed conflict

AI-enabled surveillance alters the conduct of hostilities by accelerating decision-making and expanding the scale of monitoring. In conflict settings, algorithmic systems may be used to identify potential targets, classify individuals as security threats or predict future behaviour. In non-international armed conflicts, which predominate on the African continent, such practices raise concerns under core IHL principles, including distinction, proportionality and precaution. From an ICL perspective, the use of AI-enabled surveillance may facilitate war crimes and crimes against humanity, including unlawful killings, arbitrary detention, enforced disappearance and persecution under the Rome Statute. The automation of surveillance functions also risks reducing meaningful human judgement at critical stages of military and security operations.

Structural risks in the African context

Three factors heighten the risks associated with AI-enabled surveillance in African conflicts. First, oversight mechanisms are often weak. Emergency powers, military-led counter-terrorism operations and limited judicial scrutiny restrict accountability for surveillance practices. Second, most African states rely on imported surveillance technologies supplied by foreign vendors.² Procurement processes are frequently opaque and rarely include robust human rights due diligence. Third, accountability pathways are fragmented. Domestic prosecutions for international crimes remain limited, regional accountability mechanisms are underutilised, and cooperation with international courts is politically sensitive in several contexts. These conditions create an environment in which technologically mediated violations may occur without effective legal scrutiny.

Existing AU norms

The AU has adopted several instruments relevant to digital governance. The African Union Convention on Cyber Security and Personal Data Protection (the 'Malabo Convention') establishes obligations relating to data protection, privacy and cybersecurity. In addition, the AU adopted a Continental Artificial Intelligence Strategy in 2024, emphasising ethical and inclusive AI development. While these instruments represent important normative progress, they are not designed for armed conflict. They do not address military or security uses of AI, do not integrate IHL standards and do not clarify criminal responsibility for AI-enabled harm. Moreover, the Malabo Convention has low ratification levels, limiting its practical impact.

Surveillance and international criminal responsibility

Surveillance as a facilitator of international crimes

AI-enabled surveillance may facilitate the commission of international crimes by enabling identification and targeting of individuals or groups. Pattern-based or discriminatory surveillance may contribute to collective punishment or persecution, potentially amounting to war crimes or crimes against humanity. Such conduct triggers individual criminal responsibility and may also trigger command responsibility where superiors knew, or should have known, of the risk of crimes and failed to prevent or punish them.

Surveillance as evidence

Surveillance data increasingly features in investigations, including intercepted communications, biometric records and aerial imagery. While such material may corroborate witness testimony or establish patterns of conduct, it presents evidentiary challenges. Practitioners must assess authenticity, chain of custody and the legality of data collection. In some cases, surveillance evidence may itself have been obtained in violation of international human rights law. Many domestic courts in Africa lack the technical capacity to address these issues consistently.

Attribution and the use of AI systems

Attribution of responsibility remains a central challenge. AI systems distribute decision-making across political authorities, military commanders, operators and private suppliers. This diffusion has led to arguments that algorithmic complexity undermines findings of intent or control. ICL does not recognise algorithmic responsibility. ICC jurisprudence confirms that the use of complex systems does not exclude liability where the accused was aware of the risk of criminal outcomes arising from their conduct. Nonetheless, the absence of regional guidance creates uncertainty and inconsistent practice.

Institutional constraints within the AU

The AU’s limited engagement with AI in armed conflict reflects institutional and political constraints. Member states are cautious about regulation perceived to affect national security. Technical expertise within AU organs remains limited. Responsibility for AI governance, peace and security, and human rights is dispersed across multiple institutions, with limited coordination.³ These constraints explain the AU’s incremental approach but underscore the need for targeted regulatory development.

Regulatory options

Several practical steps could strengthen the AU’s response.

Guiding principles on AI and armed conflict

The Malabo Protocol with its technology-adjacent provisions that demonstrate the drafters’ anticipated new forms of harm, could be interpreted as addressing corporate liability for AI systems and command responsibility for international crimes. However, where interpretive shortfalls arise, the Malabo Protocol can have amendments or supplementary instruments to address AI effectively. Therefore, the AU could adopt non-binding guiding principles addressing AI-enabled surveillance in armed conflict. These could affirm the continued applicability of IHL and ICL, require meaningful human control over surveillance-based targeting and discourage mass or discriminatory surveillance in conflict settings. AU soft-law instruments have previously influenced state practice.

Integration into peace and security frameworks

AI-related risks could be incorporated into the African Peace and Security Architecture, including early warning systems and peace support operations. The Peace and Security Council could explicitly consider surveillance technologies when authorising missions.

Clarification of command responsibility

The AU could issue interpretive guidance reaffirming that reliance on AI systems does not displace command responsibility. Obligations to prevent, repress and punish international crimes remain unchanged.

Corporate DUE DILIGence

Regional standards on human rights due diligence for companies supplying surveillance technologies to conflict-affected states would align with the United Nations Guiding Principles on Business and Human Rights and emerging jurisprudence on aiding and abetting liability.

Capacity building

Judges, prosecutors and investigators require training on AI systems and digital evidence. Collaboration between the AU, the International Bar Association and international partners could support judicial preparedness.

Conclusion

AI-enabled surveillance technologies are already shaping armed conflicts in Africa. The absence of conflict-specific regional regulation creates risks for civilian protection and accountability. While the AU faces political and institutional constraints, targeted regulatory measures grounded in existing legal frameworks are feasible. For war crimes practitioners, AI will increasingly feature in questions of evidence, attribution and responsibility. Ensuring that legal standards keep pace with technological developments is essential to preserving accountability in contemporary armed conflict.

Notes

1      African Union, AU Data Policy Framework, available at: https://au.int/sites/default/files/documents/42078-doc-DATA-POLICY-FRAMEWORKS-2024-ENG-V2.pdf; African Union, The Digital Transformation Strategy 2020-2030, available at: https://au.int/sites/default/files/newsevents/workingdocuments/37470-wd-annex_2_draft_digital_transformation_strategy_for_africa.pdf. Both accessed 2 June 2026.

2      Anand Sheombar and Sebastian K Skelton, ‘Who Supplies Digital Surveillance Technologies to African Governments?’ in Tony Roberts and Admire Mare (eds), Digital Surveillance in Africa: Power, Agency and Rights (Zed Books 2025), 183–210.

3      Benjamin O Ajah and Tshimangadzo O Magadze, ‘The African Union and Organised Crime: Between Normative Frameworks and Practical Constraints’ (2025) 14(3) Journal of African Union Studies 23.