Autonomous driving won’t be 100% safe: could be 10x safer than human drivers this decade

Show summary Hide summary

Tony Han, CEO of autonomous driving firm WeRide, compares today’s rush to perfect driverless cars to the earliest days of flight — a slow, risky climb toward a transformative breakthrough. His comments, made at a major forum in Riyadh, capture the tension between public concern over crashes and the industry’s push toward safer, scalable mobility.

Why pioneers accept setbacks: the Wright brothers analogy and the case for bold innovation

Han told attendees that incremental failures are part of technological progress. He pointed to history to argue that early mishaps grab headlines but ultimately lead to safer systems.

His message: entrepreneurs must be willing to test, fail, and refine in public to reach breakthroughs that will benefit future generations.

Crash figures, injury rates and what the data show about safety

Autonomous vehicles have been involved in a number of well-publicized incidents in recent years, raising questions about safety compared with human drivers.

  • Reported incidents from 2019–2024 reached nearly 4,000, with about one in ten resulting in injuries.
  • Industry comparisons show traditional cars had roughly 4.1 accidents per million miles in 2021.
  • By contrast, automated vehicles registered roughly 9.1 accidents per million miles in that period, according to legal analyses.

Despite a higher accident frequency per mile, some metrics point to fewer severe outcomes when automation is involved. Recent independent analyses suggest a large drop in certain types of claims for driverless fleets.

Insurance data and early safety signals

A commissioned review of liability claims found that driverless platforms reported far fewer injury and property claims than human-driven vehicles. Specifically, the insurer’s analysis showed dramatic percentage reductions in both bodily injury and property damage claims for autonomous systems.

Analysts say widespread adoption and better systems could also lower insurance costs dramatically over time, cutting per-mile liability costs by more than half across a 15-year horizon.

Commercial rollout: WeRide’s expansion and the global robotaxi push

WeRide has been expanding beyond China, securing permissions to operate in multiple markets and partnering to launch services in new cities.

  • Commercial robotaxi deployments in the U.S. numbered in the low thousands, with forecasts showing sharp growth.
  • WeRide announced plans to offer robotaxi rides in Riyadh through a major ride-hailing partner.
  • The company holds driving permits across several countries, enabling broader testing and commercial services.

Market projections from consulting firms estimate the autonomous vehicle ecosystem could generate hundreds of billions in global revenue within a decade.

Regulation, city adoption and the path to mainstream use

Governments and city authorities are actively working to integrate autonomous vehicles into urban transport. Han said regulators in regions such as the Middle East and parts of North America have been receptive to pilot programs and rulemaking.

San Francisco, for instance, has seen a notable rise in commercial autonomous vehicles on its streets, reflecting rapid local adoption by some operators.

Technology milestones: from partial automation to the promise of L5

Autonomy is categorized by levels, from L1 (driver assistance) up to L5 (full autonomy, no human input). Today’s commercial fleets typically operate between L2 and L4.

  • L2–L4 systems share driving tasks with humans or handle most scenarios under supervision.
  • L5 promises fully driverless operation across all conditions.

Han described true L5 capability as a watershed moment for the industry — a technological leap comparable to major AI advances in other fields. He expects L5 performance to match top human drivers within the next decade, enabling wider deployment.

Technical hurdles and industry focus areas

Achieving L5 requires breakthroughs in perception, decision-making and robustness across weather and traffic conditions. Companies are also refining redundancy, testing edge cases, and improving real-world validation.

  • Sensor fusion and real-time mapping
  • Edge-case simulation and scenario coverage
  • Redundant control systems for safety

Balancing risk and reward: industry leaders on safety claims

Executives maintain that while no new technology is perfect, automated systems can reduce harm overall. Han emphasized that the industry’s aim is not perfect safety, but meaningful improvement over human-driven travel.

His position: with ongoing iteration and scale, driverless vehicles can be substantially safer than average human drivers.

Give your feedback

★★★★★

Be the first to rate this post
or leave a detailed review


Post a comment

Publish a comment