AVs vs Humans: Safety Comparisons Need a Standard

Why It Matters

Without a standardized safety benchmark, consumers and policymakers cannot reliably assess AV risk, hindering adoption and regulatory approval. Consistent metrics are essential for insurance pricing, liability allocation, and industry credibility.

Key Takeaways

• Waymo reports tenfold fewer serious injury crashes vs humans
• No universal “average driver” benchmark exists across regions
• Swiss Re analysis shows 88% drop property‑damage claims
• Industry adopts Safety Case frameworks with CAE methodology
• Regulators consider independent safety verification to boost trust

Pulse Analysis

The surge in autonomous‑vehicle mileage has produced a flood of company‑sourced safety statistics, yet the industry still lacks a common yardstick for measuring performance against human drivers. Waymo’s publicly released figures—ten times fewer serious‑injury incidents and an 88% reduction in property‑damage claims—illustrate the potential upside, but they are anchored to a loosely defined "average driver" baseline that shifts with local traffic patterns, driver skill levels, and regulatory environments. This ambiguity makes it difficult for insurers, legislators, and the public to gauge true safety improvements.

Compounding the problem is the absence of a universally accepted human‑driver standard. In the United Kingdom, the legal notion of a "careful and competent driver" serves as a reference point, yet even that definition varies when applied to professional taxi operators, heavy‑goods vehicle holders, or everyday commuters. Stakeholders across the United States, China and Europe echo this concern, urging a more nuanced approach that blends naturalistic driving data, scenario‑based testing and rigorous simulation. Independent, agency‑led audits are being proposed to counterbalance the inherent bias of self‑reported data and to build confidence among skeptical consumers.

In response, the sector is coalescing around Safety Case frameworks, such as Tensor’s Claims‑Arguments‑Evidence (CAE) model, which articulate safety claims, logical arguments and verifiable evidence. These multi‑layered assessments move beyond single‑metric comparisons, integrating normalized crash rates, disengagement statistics, root‑cause taxonomies and functional‑safety principles. As governments contemplate mandatory safety verification, the convergence on standardized safety cases could become the de‑facto metric for AV certification, paving the way for broader market acceptance and more predictable insurance underwriting.