As artificial intelligence (AI) systems, particularly machine learning (ML)-based solutions, become increasingly prevalent in the automotive sector, ensuring their safe and reliable deployment is critical. ISO23894 [1] provides essential guidelines for risk management in AI systems, helping organizations mitigate potential hazards associated with these technologies. However, the automotive industry requires more specific frameworks to address the unique challenges of integrating machine learning into safety-critical systems from both process and product perspectives. This paper [15] presents a new, comprehensive risk management process designed for machine learning applications in automotive systems, building on the principles of ISO23894 while aligning with the new requirements introduced by A-SPICE (Automotive SPICE) Process Assessment Model (PAM) version 4.0 for machine learning. The proposed process incorporates AI-specific risk mitigation strategies and integrates them into the existing automotive software and systems development lifecycle, focusing on aspects such as model explainability, validation, traceability, and continuous monitoring. By aligning ISO23894’s guidance with the updated A-SPICE PAM V4.0, this paper provides a robust framework to improve the safety, performance, and compliance of ML-driven automotive systems, and provides and improvement proposal to the A-SPICE for machine learning as a reference.