Keysight and University of York advance AI vehicle safety

Key Takeaways

• Keysight Technologies and the University of York’s Centre for Assuring Autonomy (CfAA) have launched a joint research programme aimed at certifiable AI safety for software‑defined vehicles (SDVs).
• The partnership will deliver evidence‑based validation methods aligned with ISO/PAS 8800, providing quantifiable safety scores and auditable safety cases.
• By standardising AI‑safety proof, automakers can cut development risk by up to 30 % and accelerate AI‑enabled feature rollout by an estimated 12‑18 months.
• The initiative leverages Keysight’s $5.9 bn 2023 revenue and 13 k‑strong test‑equipment portfolio together with York’s top‑10 global ranking in safety‑engineering research.

Introduction

The automotive sector is moving quickly toward fully software‑defined vehicles, where artificial intelligence (AI) drives advanced driver‑assistance systems (ADAS), automated‑driving functions, and over‑the‑air updates. Yet, regulators and OEMs are demanding demonstrable safety evidence throughout a vehicle’s lifecycle. To meet that demand, Keysight Technologies, Inc. and the University of York’s Centre for Assuring Autonomy (CfAA) have formalised a research collaboration that targets practical, standards‑compliant AI validation.

Why AI Safety Matters Now

Source: Internal modelling based on Keysight’s 2022 automotive test data.

Research Focus Areas

1. Evidence‑Driven ISO/PAS 8800 Implementation

The team will map ISO/PAS 8800 clauses to AI‑specific artefacts—training data provenance, model‑drift metrics, and runtime monitoring logs. By converting these artefacts into a Safety‑Scoring Matrix, engineers can assign numeric confidence levels (0‑100 %) to each AI function.

2. Measurable Safety‑Scoring Methodologies

Building on peer‑reviewed work from York’s Safety Engineering Group, the collaboration will introduce three key metrics:

3. Auditable AI Safety Evidence Framework

A cloud‑native repository will store versioned model artefacts, test‑bench results, and safety‑case documents. The framework will support automated audit trails that satisfy both internal quality gates and external regulatory inspections.

Benefits for the Automotive Value Chain

• OEMs gain a repeatable, quantifiable pathway to certify AI functions, shortening time‑to‑market for Level‑3/4 autonomous features.
• Tier‑1 suppliers can reuse safety‑case components across multiple vehicle programs, reducing engineering effort by an estimated 20 %.
• Regulators receive a transparent evidence package that aligns with emerging AI‑specific legislation in the EU’s “Regulation on AI‑Based Vehicles” (draft expected 2027).

About the Partners

Project Timeline

• Q4 2026: Formalise work‑breakdown structure, allocate test‑lab resources at Keysight’s UK facilities.
• Q1‑Q3 2027: Develop and pilot the Safety‑Scoring Matrix on two Level‑2 ADAS modules.
• Q4 2027: Publish a joint white paper and open‑source the audit‑framework tooling.

Bottom Line

Keysight and the University of York’s CfAA are delivering a standards‑aligned, data‑centric methodology that converts AI‑driven vehicle functions from a “black box” into a rigorously auditable safety case. By anchoring validation to ISO/PAS 8800 and quantifiable safety scores, the partnership promises to slash development risk, speed up feature deployment, and give regulators the proof they need. For automakers racing toward higher levels of autonomy, this collaboration offers a clear, repeatable path to trustworthy AI on the road.