The threat model for a Level 4 autonomous vehicle isn't just adversarial attack surfaces - it's the entire stack's ability to make auditable, safety-critical decisions in unpredictable urban environments. MOTOR Ai is a Berlin-based company engineering certified autonomous driving software for public transport, where the stakes aren't theoretical. Their architecture merges perception and decision-making using Active Inference, a framework drawn from cognitive neuroscience, processing multi-sensor fusion inputs to generate deductive decisions in complex traffic scenarios.
The software is designed around explainability and regulatory compliance from the ground up, aligning with European safety standards. A modular structure includes redundant compute, notably a dedicated Minimal Risk Maneuver (MRM) computer as a fail-safe layer. This isn't a Silicon Valley move-fast play - it's German engineering applied to vehicle autonomy with certification as a first-class requirement, not an afterthought.
The technical domains run deep: autonomous driving, multi-sensor fusion, perception, and decision-making systems all converging on a platform built for public transport and autonomous mobility deployments. For anyone working at the intersection of safety-critical software and security, the challenge here is ensuring that a system designed for explainability and certification can also withstand the adversarial realities of operating on public roads.