At the fourth AI-Defined Vehicle Forum held on September 17, Xiang Wenling, marketing director of Neusoft Reach (a subsidiary of Neusoft), outlined the company's vision for AI operating systems in vehicles. She argued that as the industry shifts from software-defined to AI-defined vehicles, the operating system must evolve beyond being a mere host for AI applications to become a unified manager of models, data, computing power, and AI agents. According to Xiang, the deeper AI integrates into the vehicle, the higher the demands on the underlying software's stability, security, and standardization.
"AI OS needs a more solid foundation," Xiang said, emphasizing that while AI capabilities can iterate rapidly, the operating system must prioritize safe, reliable, and real-time vehicle operation. To address this, Neusoft Reach is advancing its NeuSAR OS suite, which includes products like NeuSAR cCore and NeuSAR aCore based on AUTOSAR standards, providing foundational software for various controllers and computing platforms. The company's basic software has already been adapted to over 100 mainstream automotive-grade chips from domestic and international suppliers, accumulating extensive mass-production and reliability verification experience, according to the Chinese outlet D1EV.
Three Stages of AI OS Evolution
Neusoft Reach envisions AI OS evolution in three stages: "AI in OS," "AI for OS," and "AI as OS." In the first stage, AI enters the car as individual functions or applications. In the second, the operating system provides foundational capabilities like data, models, and computing power for AI. In the final stage, AI becomes deeply embedded in the vehicle's operating system, with agents not just as standalone apps but able to directly call upon services across different vehicle domains.
To support this, the company has built the NeuSAR Service Framework to enable service-oriented architecture (SOA) capabilities across the vehicle, and an AI Framework for unified management of data, models, and computing resources, providing a foundation for AI agents to access vehicle functions. Xiang noted that Neusoft Reach has already integrated AI into software development processes such as requirements analysis, configuration, coding, test case generation, and fault diagnosis. With large models and intelligent development tools, tasks like API retrieval, code assistance, problem localization, and test script generation can be automated, shortening development and testing cycles.
Xiang stressed that AI-defined vehicles are not just about adding a large model to the cockpit or enhancing driver assistance. As AI penetrates deeper into the vehicle architecture, the entire automotive software system needs to be redesigned. For automakers, the challenge is not just "how to put AI into the car," but how to balance rapidly evolving AI capabilities with the industry's long-standing standards for safety, reliability, and mass production. "AI can get faster, but the foundation that supports it in the car cannot become more fragile," she said.
