Investigating Artificial Intelligence Digital Sovereignty in Mobile Shopping Apps: A Case Study of Nigeria
Source: arXiv:2608.06364 · Published 2026-08-06 · By George Grispos, Sajda Qureshi
TL;DR
This paper investigates the impact of Artificial Intelligence (AI) on individual digital sovereignty within Nigerian mobile e-commerce applications, focusing on the relationship between AI transparency and user control. The study addresses the growing use of AI-driven personalization, recommendation, and conversational agents embedded in popular Nigerian shopping apps (Jumia, Konga, and Jiji), amid a context of rapid digital adoption alongside socio-economic challenges such as fraud and limited user awareness. Using an interpretive methodology that combines forensic analysis of Android app artifacts with contextual document analysis, the authors provide empirical evidence that AI functionalities are widely integrated, yet disclosures around AI usage remain scarce or buried within complex legal documents. Nigeria’s moderate levels of AI awareness and trust compound concerns about individual users’ ability to meaningfully control their data and interactions, a key element of digital sovereignty.
Key findings
- Forensic analysis identified AI-related artifacts in all three apps (Jumia, Konga, Jiji), including personalization SDKs like Netcore Smartech (Konga) and MoEngage (Jumia), as well as conversational AI integrations (e.g., Sprinklr chatbot in Jumia and Intercom messaging in Jiji).
- AI-related user data stored on-device includes search histories, product views, session identifiers, and messaging telemetry, indicating AI-driven engagement and recommendation systems.
- Transparency analysis revealed no explicit AI disclosures in Google Play Store descriptions for any app; only Jiji’s privacy policy explicitly acknowledges AI-based automated processing.
- Konga, despite forensic evidence of AI use, showed no public disclosure of AI or algorithmic decision-making in app listing or privacy policy.
- Survey data from the 2025 UN Global Survey on AI shows moderate AI awareness in Nigeria where 40% know 'a little' about AI, and 22% have no knowledge, while 41% never knowingly interacted with AI systems.
- Nigeria’s consumer digital platform dependence is high with over 70% of e-commerce done via mobile apps, but public trust in AI is low, with 52% expressing limited or no confidence.
- AI-driven personalization and profiling raise concerns about digital sovereignty through limited transparency, data control, and potential for surveillance capitalism effects.
- Server-dependent AI architectures dominate, meaning personal data is processed externally rather than on-device, increasing control risks.
Threat model
The adversary model centers on typical Nigerian mobile e-commerce consumers exposed to AI-driven personalization and automated decision-making without explicit transparency or control. Threats include loss of data control, manipulation through opaque algorithmic profiling, and exploitation by fraudulent actors leveraging AI functionalities embedded in apps. The study assumes adversaries do not have privileged access to app internals but may exploit platform opaqueness and user unawareness. The research does not target active attacker behaviors such as reverse engineering or adversarial machine learning attacks.
Methodology — deep read
The research applies an interpretive approach combining forensic analysis of mobile app artifacts with contextual document analysis to assess AI transparency and digital sovereignty implications. The threat model centers on typical consumer users with limited AI knowledge, interacting with mobile apps embedding AI features without clear disclosure or control options. Adversarial threats such as fraud are acknowledged as background risks influencing user trust.
Data includes three leading Nigerian e-commerce Android applications (Jumia, Konga, Jiji) selected due to market prominence and install bases (100M+, 10M+, 1M+ respectively). Apps were installed on a rooted Android Virtual Device (AVD) to gain filesystem access without using real user accounts or personal data, ensuring controlled forensic extraction using Android Debug Bridge (ADB). AVD rooting allowed retrieval of app directories including databases, XML files, cache, logs, and metadata.
The forensic pipeline screened SQLite databases and preference files for AI-indicative components such as SDK identifiers (Netcore Smartech, Firebase Remote Config, MoEngage, Sprinklr chatbot, Intercom messaging, Google ML Kit). Artifacts were linked to AI-driven functionalities: recommender systems, conversational agents, behavioral analytics. AI presence was inferred from SDK names, config files, and stored user interaction data rather than direct model execution.
Separately, a document analysis reviewed Google Play Store listings, privacy policies, and official platforms to assess algorithmic transparency and explicit AI disclosure, focusing on user-accessible information.
Contextual data incorporated the 2025 UN Global Survey on AI and Human Development Index to situate findings within Nigeria’s socio-economic and technological awareness landscape.
For example, Jumia’s jumia.db contained tables with recent search/view histories; LiveChatConfigEntity showed Sprinklr AI bot integration; MoEngage SDK artifacts stored user engagement metadata. Konga’s NCSmartech database and Firebase config files mapped to personalization features. Jiji’s caches and preference files evidenced Intercom and Google ML Kit use.
All analysis is reproducible via AVD emulation and rooted extraction; code/scripts for data extraction were not disclosed. No live user data was collected, and model execution or runtime AI behavior was not observed directly—AI integration was inferred.
Evaluation combined qualitative artifact interpretation with cross-referencing to public SDK documentation. AI transparency was assessed via manual review of publicly available disclosures. The study generalizes pattern insights to mobile-first emerging markets with similar digital economy and governance contexts rather than statistical inference.
Technical innovations
- Use of forensic artifact analysis in a rooted Android Virtual Device to systematically identify and classify AI-related components within mobile e-commerce apps.
- Application of an interpretive mixed-methods approach combining technical artifact recovery with contextual socio-economic data to assess digital sovereignty.
- Novel focus on AI transparency in emerging market mobile apps as a practical indicator of individual digital sovereignty.
- Classification of AI features into locally executed versus server-dependent systems based on forensic evidence and SDK characteristics.
Datasets
- Jumia app artifacts — multiple SQLite DBs and XML files extracted from rooted Android emulator — private forensic dataset.
- Konga app artifacts — multi-database extracts from rooted Android emulator — private forensic dataset.
- Jiji app artifacts — SQLite DBs, JSON caches, and shared preferences extracted from rooted Android emulator — private forensic dataset.
- 2025 UN Global Survey on AI and Human Development — public survey data on AI familiarity, usage, and trust in Nigeria.
- UN Human Development Index — public country-level socio-economic indicators.
Figures from the paper
Figures are reproduced from the source paper for academic discussion. Original copyright: the paper authors. See arXiv:2608.06364.
Fig 4: Jumia and MoEngage Engagement
Limitations
- Study limited to three Nigerian mobile apps, restricting statistical generalization beyond observed ecosystem.
- AI presence inferred indirectly from forensic artifacts and SDK references rather than direct observation of runtime model execution or decision outputs.
- Transparency assessment confined to public documentation, potentially missing non-public user disclosures or in-app explanations.
- No behavioral or black-box testing conducted to measure how AI outputs (recommendations, chat responses) adapt or affect user experience.
- Contextual socio-economic data from surveys may not fully represent current or localized user understanding and trust levels.
- Lack of adversarial evaluation to assess potential abuses from AI integration or data leakage.
Open questions / follow-ons
- How can mobile e-commerce applications better implement transparent AI disclosures to enhance digital sovereignty without sacrificing usability?
- What technical mechanisms can enforce user consent and control over AI-driven data processing on mobile platforms, particularly in emerging market contexts?
- How do live AI model behaviors and recommendation dynamics in Nigerian apps influence user trust and decision-making in practice?
- Can combined forensic and behavioral analyses scale to larger samples of apps across Africa to model regional variations in AI transparency and digital sovereignty?
Why it matters for bot defense
For bot-defense and CAPTCHA practitioners, this study underscores the complexities of AI integration within mobile platforms used in emerging markets—environments where user control and transparency are limited, yet AI-driven personalization and behavioral profiling are pervasive. Understanding how popular apps embed AI without clear user disclosure points to potential risks around consent, data sovereignty, and automated decision-making opacity. From a bot-defense perspective, these opaque AI layers may complicate reliable user-interaction modeling and elevating detection challenges, as AI personalization modifies user behavior patterns. Furthermore, the predominance of server-dependent AI processing highlights potential attack surfaces connected to data transmission and third-party SDKs.
Overall, the work signals the importance of designing bot detection and CAPTCHA mechanisms that consider not only technical robustness but also transparency and user trust in AI systems, especially in mobile-first economies with evolving regulatory frameworks. It also suggests future solutions may benefit from integrating clearer user consent and auditability into AI-driven mobile commerce scenarios to support individual digital sovereignty.
Cite
@article{arxiv2608_06364,
title={ Investigating Artificial Intelligence Digital Sovereignty in Mobile Shopping Apps: A Case Study of Nigeria },
author={ George Grispos and Sajda Qureshi },
journal={arXiv preprint arXiv:2608.06364},
year={ 2026 },
url={https://arxiv.org/abs/2608.06364}
}