2025 Automotive report highlights code quality, AI and safety
The latest survey of automotive software developers shows the focus shifting to safety alongside increased use of AI tools.
The 2025 State of Automotive Software Development Report from Perforce shows an increase in the use of AI with continued challenges around code complexity, and a larger emphasis on safety than security.
Safety was the top concern in AI vehicle development for 49% of 650 developers surveyed, so teams who are guided by functional safety standards need to employ additional considerations when using AI, as the algorithms tend to be non-deterministic. However many more engineers are using the latest ISO8800 AI guidelines for development.
“Automotive software development is becoming more dependent on AI systems, but the software must still be developed, deployed, and maintained with safety in mind,” said Perforce Director of Compliance Jill Britton. “AI systems bring additional challenges to achieving functional safety and to provide guidance, new and updated standards are emerging for their use in safety-critical applications.”
AI is driving autonomous vehicle design for 42% of automotive professionals (up 9% from last year) and is impacting at least some components in connected vehicles (41%). Advanced Driver Assistance Systems (ADAS) were the leading components with AI/ML applications, alongside In-Vehicle Infotainment (IVI) systems and Light Detection and Ranging (LiDAR) components.
In the previous report, security rose as a concern over safety, but with the rapid introduction of AI/ML in connected and autonomous vehicle development and design, safety is once again of higher concern in 2025.
“It depends on how they are using AI. There are three different areas,” Britton told eeNews Europe.
“The first one is a tool with AI to check code and we have no problem with that. The second one is using AI in the development of the product. There are concerns as you want to check what you get out at the end. The big problem is if you have a learning machine within the product as its non deterministic and the current functional safety standards are based on a system being deterministic.
“That’s where we are. We know that we can’t just rely on AI, it depends on the underlying data.”
“Functional safety is being turned on its head. Making sure that it doesn’t do anything that you don’t expect. ISO26262 in automotive is deterministic, with one entry point and one exit point, We then have the new generation of functional safety stafety standards that look at the risk.”
She points to the ISO8800 specification that is looking at how risk can be reduced rather than step by step. The report found that 71% of respondents are adopting ISO/DPAS 8800 for AI functional safety assurance.
“Yes, AI can be functionally safe if you apply the concept of the requirements, what are you expecting this piece of software to do and is it still fulfilling the requirements and that’s the difficult part. That turns things on their head,” she said.
Automotive software professionals are increasingly aware that maintaining high-quality code contributes to both the safety and security of the software system. But the complexity of the code base can make producing quality code challenging, especially for engineers with less than three years of experience: 57% of respondents with less than one year and 45% with one to three years expressed code complexity as their top concern; as opposed to those with more than five years (37%) who cited testing resources as their top quality concern.
While wider market conditions and challenges like the global economy and remaining competitive are driving most organizations, a consistent trend throughout the report showed an emphasis on maximizing existing resources (49%) and educating existing talent (42%).
Developers are turning to static analysis tools to manage complex code bases, ensuring compliance with industry standards like MISRA and ISO 21434. The report found that 30% of teams prioritize software quality improvements through static analysis, version control, and continuous testing tools.
The report found that 86% of respondents are using at least one coding standard, which is important for code quality with 53% using static analysis/SAST tools and 30% citing the primary reason as improving software quality.
89% are required to track code quality metrics to reduce errors — an increase of 12%, while EV software development is stabilizing, with 47% working extensively on EV systems.
The annual study was conducted in collaboration with Automotive IQ and the Eclipse Foundation and is available at www.perforce.com/resources/sca/2025-state-automotive-software-development-report.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
