Keysight and University of York Advance AI Safety for Software-Defined Vehicles
Keysight Technologies, Inc. and the Centre for Assuring Autonomy (CfAA) at the University of York in the United Kingdom have announced a new research collaboration aimed at strengthening the safety assurance of artificial intelligence (AI) technologies used in software-defined vehicles (SDVs).
The partnership brings together Keysight’s expertise in AI validation and automotive testing with the University of York’s research capabilities in systems safety and autonomous-system assurance. The organizations intend to develop practical, evidence-based approaches that can help automotive manufacturers and suppliers demonstrate that AI-powered vehicle functions are safe, reliable and fit for deployment.
As vehicles increasingly depend on software and artificial intelligence, proving the safety of these technologies has become a critical challenge for the automotive industry. Advanced driver assistance systems (ADAS), automated driving functions and other intelligent vehicle technologies are becoming more sophisticated, creating new requirements for testing, validation and safety assurance.
Unlike many traditional automotive systems, AI-based functions can involve complex algorithms and data-driven behavior. This creates additional challenges for engineers seeking to establish confidence in how an AI system will behave across different driving conditions and throughout the vehicle’s operational lifecycle. The collaboration between Keysight and the CfAA is intended to address some of these challenges by connecting academic research with practical engineering methods.
Addressing the Challenge of AI Safety Assurance
The rapid adoption of AI in automotive applications is changing how vehicle systems are developed and validated. AI technologies can support perception, decision-making, driver assistance and automated driving capabilities, but their increasing role also means manufacturers must provide stronger evidence that these systems perform safely.
For automotive companies, demonstrating AI safety is not limited to determining whether a system works under normal operating conditions. Engineering teams also need to consider how systems perform in unusual situations, how they respond to changing environments and how their behavior can be evaluated consistently.
The new collaboration will therefore focus on methods for validating AI systems and producing structured evidence that can be used to support safety arguments.
A key objective is to help bridge the gap between theoretical AI safety principles and the practical requirements faced by automotive engineering organizations. By developing repeatable methodologies, the partners aim to make it easier for teams to understand, document and communicate the safety of AI-enabled vehicle functions.
The work could ultimately support manufacturers and suppliers as they address emerging industry expectations and standards while attempting to reduce development risks and accelerate the introduction of AI-enabled technologies.
Supporting Software-Defined Vehicle Development
Software-defined vehicles represent a major shift in automotive engineering. Instead of relying primarily on fixed hardware functions, SDVs increasingly use software to deliver, update and enhance vehicle capabilities.
This model can provide manufacturers with greater flexibility and enable new functions to be introduced during a vehicle’s lifecycle. However, it also increases the importance of software validation and continuous safety assurance.
AI adds another layer of complexity. Automotive organizations must be able to establish confidence not only in the underlying software but also in the AI components that may influence system behavior.
The Keysight-CfAA collaboration is designed to help address this challenge by developing approaches for structured AI safety cases. Such safety cases can bring together arguments, assumptions and supporting evidence to demonstrate that an AI-based function satisfies defined safety objectives.
Evidence-driven assurance can also help engineering teams make their safety activities more auditable. This can be particularly important in safety-critical automotive applications, where organizations need to demonstrate that appropriate processes and validation activities have been followed.
Focus on ISO/PAS 8800
One of the central areas of research will be the implementation of requirements associated with ISO/PAS 8800, an automotive standard focused on the safety of artificial intelligence.
The collaboration will explore ways to translate relevant requirements into measurable and practical methodologies that can be used by automotive engineering teams.
The research is expected to examine safety-scoring approaches that are grounded in both academic research and established industry standards. Developing measurable approaches could help organizations evaluate AI safety more consistently and provide clearer evidence to support their safety claims.
Another important objective is the development of practical frameworks for producing auditable AI safety evidence.
For manufacturers and suppliers, having a structured approach to evidence generation could simplify the process of demonstrating how AI systems have been evaluated. It may also make it easier to identify gaps in validation activities and determine what additional evidence is required before a system is deployed.
Combining Academic Research With Engineering Expertise
The University of York’s Centre for Assuring Autonomy brings academic expertise in safety engineering and the assurance of complex autonomous systems.
The CfAA has conducted research into the assurance of autonomous technologies across multiple sectors, including transportation. Its work focuses on developing approaches that can help organizations understand and manage the safety challenges associated with increasingly autonomous systems.
Keysight, meanwhile, provides testing, measurement and validation technologies used across the electronics and automotive industries. Its work in AI validation is focused on helping organizations evaluate AI-enabled systems and establish confidence in their performance.
By combining these capabilities, the organizations aim to develop methodologies that can be applied to real-world automotive engineering environments.
The partnership is particularly focused on making research useful for engineers who must transform high-level safety principles into concrete development and validation activities.
Simon Burton, Chair in Systems Safety at the University of York, said the automotive industry is reaching an important stage as AI becomes increasingly integrated into vehicle functionality.
According to Burton, robust, evidence-based evaluation approaches are essential as automotive companies adopt these technologies. He also highlighted the role of the Centre for Assuring Autonomy in developing frameworks and guidance that can be used by safety professionals in the transportation sector.
The collaboration is intended to further support the development of practical methods that translate AI safety principles into engineering practices suitable for safety-critical environments.
Keysight’s AI Validation Approach
Keysight’s contribution to the collaboration will include its AI validation capabilities and broader experience in automotive technology development.
Lukas Klose, Head of the Automotive AI Solution Center at Keysight, said automotive organizations need approaches that are both practical and scalable when developing confidence in AI-enabled systems.
The partnership is expected to combine the University of York’s research in safety assurance with Keysight’s holistic AI Validation Framework. The objective is to help engineering teams develop structured evidence that can support AI safety cases.
This approach could be particularly valuable as automotive organizations seek to deploy AI technologies while maintaining alignment with international standards.
Rather than treating AI validation as a single activity performed at the end of development, the collaboration emphasizes evidence generation and assurance across the product lifecycle. This reflects the increasingly continuous nature of software-defined vehicle development, where software and AI functionality may evolve after a vehicle enters service.
Building Confidence in AI-Enabled Automotive Systems
Consumer and industry expectations around automated vehicle technology are increasing. Drivers and vehicle manufacturers need confidence that intelligent systems will behave appropriately in a wide range of situations.
For automotive developers, this requires more than conventional functional testing. AI systems may need to be evaluated against a wide range of scenarios, datasets, environmental conditions and potential edge cases.
A structured safety-assurance methodology could help organizations organize this evidence and connect individual validation results to broader safety claims.
The ability to create auditable evidence may also become increasingly important as regulatory frameworks and industry standards evolve. Automotive organizations need to be able to demonstrate not only that testing has taken place but also why the selected validation activities provide reasonable confidence in system safety.
The research partnership therefore has the potential to contribute to a more systematic approach to AI assurance in automotive development.
Potential Impact Across the Automotive Supply Chain
The benefits of improved AI safety assurance could extend beyond vehicle manufacturers.
Automotive suppliers developing AI components, software platforms, sensors, electronic control systems and automated driving technologies also need reliable methods for validating their products.
A common evidence-based approach could help suppliers communicate safety information to vehicle manufacturers and other stakeholders. It could also support greater consistency in how AI-enabled components are assessed throughout the automotive supply chain.
For original equipment manufacturers, structured evidence could assist with integrating AI technologies from multiple suppliers while maintaining an overall safety argument for the vehicle.
This becomes increasingly important as software-defined architectures allow vehicle functions to depend on interconnected systems supplied by different technology providers.
Future Development of Keysight Solutions
The research is expected to inform the future development of Keysight’s AI Software Integrity Builder.
The solution is intended to support activities related to AI validation and software integrity. Findings from the University of York collaboration could help enhance its capabilities for AI safety arguments, safety evidence generation and validation activities.
This creates a potential pathway for academic research to influence practical commercial engineering tools.
As the automotive industry moves toward more sophisticated software-defined architectures, such tools could become increasingly important for organizations seeking to integrate AI while maintaining rigorous development and assurance processes.
The collaboration between Keysight Technologies and the Centre for Assuring Autonomy reflects a broader industry shift toward evidence-based AI safety.
Artificial intelligence has the potential to transform vehicle development, enabling more advanced driver assistance, automation and software-based functionality. At the same time, the safety implications of deploying AI in vehicles require rigorous validation and assurance.
By combining academic research, safety engineering expertise and industrial AI validation capabilities, Keysight and the University of York aim to develop practical approaches that can help automotive organizations address these challenges.
The research will focus on structured AI safety cases, measurable safety methodologies, auditable evidence and alignment with emerging automotive standards such as ISO/PAS 8800. These efforts could help engineering teams better understand and document the safety of AI-enabled functions while supporting more efficient development processes.
As software-defined vehicles become increasingly central to the future of transportation, establishing confidence in their AI capabilities will remain a major priority. Partnerships that connect research with practical engineering applications could play an important role in developing the methods needed to safely deploy these technologies.
Ultimately, the collaboration seeks to help automotive manufacturers and suppliers move toward a future in which AI-powered vehicle systems are not only more capable but also supported by clear, structured and defensible evidence of their safety and reliability.
Source Link:https://www.businesswire.com/
