Safety and Regulation in Autonomous Transport

When a robot drives, who is the driver? Regulators worldwide have spent five years answering that. In 2026, we finally have harmonized answers — and a safety record to debate.

The Three Pillars of Safety

• ISO 26262 & ISO 21448 (SOTIF): Functional safety + safety of the intended function (what if AI misclassifies?)
• UL 4600: Safety case methodology — companies must prove "absence of unreasonable risk"
• IEEE 2857: Disclosure of autonomy performance data to regulators

UN-ECE WP.29 — The Global Rulebook

UN Regulation No. 157 (ALKS) now allows Level 3 up to 130 km/h, and No. 171 governs Driver Control Assistance Systems. For Level 4, national laws diverge: US (STATE patchwork + NHTSA voluntary guidance), EU (EU 2022/1426 ADS approval framework), China (Shenzhen & Beijing permit driverless with safety operator remote), and Japan (2023 amended Road Traffic Act).

Liability Shift

In Level 4 robotaxi operation, the Automated Driving System Entity — typically the fleet operator — holds liability, not the passenger. Insurance is product liability + fleet policy. Germany's StVG and UK's Automated Vehicles Act 2024 enshrine this. Data recorders (DSSAD) are now mandatory: 30 seconds before/after any event must be stored.

Vision Zero Goal

94% of crashes involve human error. Autonomy aims for zero. Yet public trust lags: 61% of Americans still say they'd be afraid to ride in a driverless car (AAA, Jan 2026). Transparency — publishing disengagement and crash rates — is how industry earns trust.