Abstract: Now, companies in industries including manufacturing, shipping, supply chain, healthcare, and critical infrastructure can attain previously unheard-of levels of automation, efficiency, and predictive capabilities because of the convergence of artificial intelligence (AI) and cyber-physical systems (CPS). But this connection brings with it serious new security issues. Cyber-physical attacks that modify digital control levels, take advantage of AI models, or interact with physical devices can seriously impair company operations, leading to monetary losses, security threats, and damage to one's reputation. The impact of such attacks on AI-enabled business systems is examined in this article through the development of a comprehensive threat vector taxonomy that covers the cyber, physical, and AI/model layers. We provide impact metrics that link technical disruptions to measurable business consequences, such as operational inefficiencies, economic costs, downtime, and fines from the government. We use real-world occurrences, benchmark CPS datasets (SWaT, WADI, BATADAL) for experimental evaluations, and controlled attack scenarios to show how vulnerable AI-driven decision-making pipelines are too adversarial and supply-chain threats. We also examine mitigation strategies like secure model lifecycle management, anomaly detection, robust machine learning, and sensor redundancy. According to the study, in order to preserve the credibility of AI-enabled business CPS, extensive defences, regulatory standards, and a strong system architecture are essential.

Keywords: Cyber-Physical Systems (CPS); AI Security; Business Systems; Adversarial Machine Learning; Cyber-Physical Attacks; Supply Chain Security; Industrial Control Systems (ICS); Anomaly Detection; Model Poisoning; Resilient AI; Critical Infrastructure Protection; Business Risk; Operational Technology (OT) Security; Secure AI Lifecycle.

Downloads: | DOI: 10.17148/IJIREEICE.2025.13925