Systemic Risk of Using AI-Generated Synthetic Data in Autonomous Vehicle Development

Funder

AISI and EPSRC

Value 

£253,000

Project Team 

Lead: Coventry University

Dr Qian Lu
Huw Davies
Vasile Palade

Collaborators: HORIBA MIRA

Duration 

05/02/2025-04/02/2026

Project overview

This project examines the systemic risks introduced by AI-generated synthetic data in autonomous vehicle (AV) development. The research explores hazards in synthetic datasets (e.g. biases, inaccuracies, or unrealism, etc) that can compromise AV performance and create cascading effects across AV fleets—potentially impacting traffic flow, safety, emergency services, and social equity. Drawing on advanced machine learning (e.g., generative models), simulation techniques such as IPG Carmaker and SUMO, and socio-demographic data, the project aims to identify and quantify these hazards. The outcome will be a comprehensive risk assessment framework—complete with standardised evaluation methods and policy recommendations—to guide industry practices and foster equitable, reliable AV deployment.

Project objectives

1. Identify, categorise, and characterise technical and social risks arising from the use of AI-generated synthetic data in AV systems.
2. Assess how identified risks affect individual AVs, entire fleets, and transportation networks, as well as societal inequalities.
3. Formulate standardised evaluation methods, metrics, and criteria to integrate risk management into existing AV development and safety standards.
4. Perform a policy gap analysis and propose data governance and regulatory guidelines to mitigate identified risks and support equitable AV integration.