DriveSafeAI: Advancing AI Safety in Automated Vehicles

DriveSafeAI is a pioneering research initiative dedicated to establishing robust, evidence-based AI safety assurance for assisted and automated vehicles (AVs). Spearheaded by Wayve and WMG’s Safe Autonomy Research Group, the project develops cutting-edge methodologies, tools, and datasets to validate AI performance in real-world driving conditions. At its core is the Safety Pool™ Scenario Database—the world’s largest public repository of test scenarios for autonomous driving—enabling transparent, scalable, and agnostic safety validation across diverse operational design domains. The project introduces the OASISS framework, a novel approach to AI safety that separates behavioural competencies from operational contexts, supporting regulatory confidence and global standardization. By leveraging Wayve’s PRISM-1 simulation techniques and aligning with international standards (ISO, UNECE, EU), DriveSafeAI strengthens the foundation for safe, reliable, and commercially viable self-driving technology.

Key Features:

• Safety Pool™ Scenario Database: The world’s largest public database of test scenarios for automated driving systems.
• OASISS Framework: A scalable, ODD-based AI safety assurance methodology for generalizable validation.
• Behavioural Competency Taxonomy: A new classification system to measure and verify driving skills in AI models.
• Edge-Case Simulation: Uses PRISM-1 simulation techniques to generate realistic and diverse edge-case scenarios.
• Real-World vs. Simulated Performance Analysis: Compares AI model behavior in simulation and real-world environments to bridge the testing gap.
• Independent Safety Validation: Provides objective, research-driven evidence to support regulators and policymakers.
• Redundant Safety Architecture Design: Based on ISO/TS 5469 reference architecture to enhance system reliability.
• Global Collaboration Platform: Promotes shared development and testing of self-driving technology across industry and academia.
• Industry & Regulatory Alignment: Aims to inform national (BSI), international (ISO), and regulatory (UNECE, EU) standards for AV safety.

Pricing: DriveSafeAI is a publicly funded research project and does not offer commercial pricing tiers. It operates under a freemium model by providing open access to its Safety Pool™ datasets and research outputs at no cost, while supporting industry adoption through collaboration and policy guidance.

Conclusion: DriveSafeAI is a transformative research effort that sets a new benchmark for AI safety in autonomous vehicles, offering unparalleled tools and data to ensure trustworthy, scalable, and globally aligned self-driving technology.