Keysight Technologies has entered a research collaboration with the Centre for Assuring Autonomy (CfAA) at the University of York to develop practical methods for validating AI systems used in software-defined vehicles. The partnership targets a growing industry problem: as AI takes on a larger role in ADAS, automated driving and SDV functions, manufacturers need repeatable ways to prove these systems are safe and reliable across the vehicle lifecycle, not just at launch.
The work will focus on three areas: building evidence-based approaches that support implementation of ISO/PAS 8800, developing measurable safety-scoring methods based on academic research and existing standards, and creating frameworks for producing auditable safety evidence. The intent is to help engineering teams construct structured AI safety cases and demonstrate justified confidence in AI behaviour more efficiently, rather than treating safety argumentation as an ad hoc exercise late in development.
CfAA brings systems-safety research and existing frameworks already used by transport-sector safety professionals, while Keysight contributes its AI validation techniques. According to the companies, findings from the collaboration are expected to feed into future updates of Keysight's AI Software Integrity Builder tool, extending its support for generating safety arguments and evidence aligned with automotive standards.
For engineers, the collaboration signals movement toward more standardised, auditable methods for justifying AI safety in production vehicles, an area where formal, cross-industry practice has so far lagged behind the pace of AI deployment in vehicle software.
Source: ADAS & Autonomous Vehicle International — read the original article