Testing autonomous vehicles and AI: perspectives and challenges from cybersecurity, transparency, robustness and fairness

AI in autonomous vehicles requires new testing approaches: Traditional testing methods are insufficient for AI-based systems in vehicles. The paper emphasizes that Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS) need specialized testing methodologies to ensure safety, moving beyond conventional automotive validation procedures.

Cybersecurity audits are critical for trustworthy AVs: The study identifies cybersecurity as an essential requirement for trustworthy AI in autonomous vehicles, highlighting that comprehensive security audits must be integrated into development and deployment processes to protect against vulnerabilities that could compromise vehicle safety.

Explainability in AI decision-making is non-negotiable: The paper stresses the importance of transparency in how AI systems make driving decisions. Users and regulators need to understand why autonomous vehicles make specific choices (such as emergency maneuvers), which is essential for accountability and public trust.

Robustness and ethical behavior must be actively assessed: The research emphasizes that predictive systems in AVs must be tested not only for technical performance but also for ethical behavior and robustness under real-world conditions, ensuring vehicles behave fairly and safely across diverse scenarios and populations.

Regulatory frameworks like the EU AI Act will reshape AV development: The paper notes that emerging regulations have significant implications for how autonomous vehicles are designed and tested, requiring multidisciplinary expertise to align AI systems with legal and ethical requirements before deployment.