The evolution of Autonomous off-highway vehicles (OHVs) has transformed mining, construction, and agriculture industries largely by increasing vehicles efficiency and safety. These vehicles operate in high dust, uneven terrain, and potential communication failure environments where vehicle safety is challenged. To guarantee vehicle safety in such situations, a robust architecture that combines AI-driven perception, fail-safe mechanisms, and conformance to many ISO standards is required. In unstructured environments, AI-driven perception, decision-making, and fail-safe mechanisms are not fully addressed by traditional safety standards like ISO26262 (road vehicles), ISO19014 (earth-moving machinery),ISO12100(Safety of machinery) and ISO25119 (agriculture). These standards mainly concentrate on the reliability of mechanical and electric/electronic systems. Additionally, emerging standards such as ISO21448 (SOTIF) used to detect and mitigate unsafe AI outputs, and ISO8800(AI safety) which focus on safety assurance for AI-based systems, offer valuable insights but requires additional adaptation. AI-driven systems are vulnerable to cyber threats that can endanger the vehicle safety. Integration of ISO21434 (cybersecurity for vehicles) with safety standards is needed to address safety failures induced by cyber threats. This paper introduces a hybrid safety framework that integrates ISO safety standards ISO26262, ISO19014, ISO12100, ISO25119, ISO21448, ISO21434 and ISO8800 with AI innovations enhance the safety and reliability of autonomous OHVs. The proposed framework makes use of sensor fusion, explainable AI for transparent decision-making, especially in safety-critical scenarios, and a real-time fail-safe mechanism to handle critical failure scenarios such as power failures, communication loss, and sensor degradation by switching to a safe state. To confirm the effectiveness of this hybrid approach, digital twin simulation software along with additional technologies in OHV applications are used. The results demonstrate significant improvements in more accurate fault detection, efficient responses, and overall system resilience, highlighting the benefits of merging AI safety techniques with established ISO standards.