Deposit Money Banks (DMBs) are embracing AI-powered intrusion and fraud detection, but there is no theory of adoption that asks if the AI is vulnerable to deception. Adversarial machine learning demonstrates how ‘tricky' a small, purposely made change to the input can cause a detector to mistake malicious activity for benign activity. The TOE framework is thus insufficiently specified for the process of adopting AI cybersecurity in high-risk industries, the argument of this paper is that this is the case. By combining three literatures, the paper spans the research domains of TOE adoption studies (2022-2026), adversarial machine learning in intrusion detection, and recent financial and AI regulations, such as the CBN (Central Bank of Nigeria) 2024 Risk Based Cybersecurity Framework and Article 15 of the EU AI Act. Building on this synthesis it builds an extended TOE model with a new construct, Adversarial Robustness Preparedness (ARP): an organisation's ability to anticipate, test, defend against and govern attacks to its AI model. The model proposes eight propositions connecting technological, organisational and environmental factors and ARP with the adoption of AI for cybersecurity in Nigerian DMBs. It points to ARP as a direct cause of adoption, a partial mediator of organisational and regulatory influences, and a moderator of the impact of perceived AI capability. A measurement and validation roadmap based on structural equation modelling is provided. In addition to a testable theoretical extension, the paper provides a vocabulary for bank boards and regulators and a research agenda for new and developing financial markets.
No takes yet. Share an insight, caveat, or question.
U. et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: