Artificial Intelligence (AI)-powered cybersecurity is increasingly adopted in complex and data-intensive environments, particularly in Internet of Things (IoT) settings, where traditional rule-based approaches struggle to cope with dynamic and unforeseen threats. While AI-based solutions demonstrate high detection accuracy, responsiveness, and operational efficiency, their evaluation remains dominated by technical performance indicators such as precision, recall, false-positive rates, and computational efficiency. This focus provides a partial view of the suitability of AI-based cybersecurity solutions, ignoring relevant issues related to energy consumption, carbon footprint, organizational impact, accountability, and regulatory compliance. To addresses this limitation, the paper adopts a conceptual and design-oriented research approach to propose a multidimensional key performance indicators (KPI)-based framework for comparing AI-driven cybersecurity solutions, considering technical, environmental, social, economic, and governance dimensions. The framework operationalises evaluation through a structured pipeline that combines three main elements: the AI-based cybersecurity solutions under evaluation, a baseline set of KPI, and expert judgements on the relative importance of the five evaluation dimensions. The framework operationalises evaluation through a structured pipeline. Heterogeneous KPIs are classified according to their direction of preference and normalised onto a common scale to ensure comparability. Normalised indicators are aggregated within each dimension to generate dimensional performance scores. Expert judgements are processed to derive dimension weights reflecting contextual priorities. Dimensional scores and weights are combined through weighted aggregation to produce both a composite sustainability-performance index and a multidimensional performance profile for each solution. The contribution of the paper lies in moving beyond accuracy-centric benchmarking by providing a transparent and replicable evaluation logic that makes trade-offs across dimensions explicit. Rather than treating sustainability as an external factor or an additional descriptive element, the framework integrates it directly into the evaluation process, enabling more balanced and context-sensitive comparison among AI-driven cybersecurity solutions. The proposed framework provides a methodological foundation for future empirical applications and supports more informed decision-making in the selection and governance of AI-based cybersecurity technologies.