Abstract
<jats:p>The research employs threat analysis techniques and adaptive model parameterization using machine learning to identify specific attacks targeting AI systems. For comprehensive cyber risk assessment, a semi-quantitative additive-multiplicative mathematical model with logistic normalization is proposed. The main result of the study is the development of an integrated architecture that dynamically adapts to data sensitivity and environmental context. Within the research, a risk calculation formula is mathematically derived and justified, integrating the probability of resource compromise, the impact scale of incidents, and the operational response context. The effectiveness of the proposed approach is validated through an experimental prototype, the testing of which demonstrated a significant reduction in mean time to detect (MTTD) and mean time to respond (MTTR), as well as a decrease in false positive rates. The practical significance of the proposed solution lies in providing organizations with a scalable approach to cyber risk management in multi-cloud ecosystems, enabling automated auditing processes, continuous compliance with international standards, and robust protection of data against emerging threats. Keywords: Compliance Security Posture Management (CSPM), Data Security Posture Management (DSPM), security automation, multi-level L1–L4 model, SOAR, policy-as-code, Prisma Cloud, AI/ML security, cloud infrastructure, MTTR.</jats:p>