The AV Safety Gap Is Not
About Accidents. It Is
About Evidence.
Understanding why safety improvement and safety acceptance are not always the same thing.
The Safety Evidence Gap Framework™ maps how operators, regulators, insurers, and the public rely on different evidence when evaluating AV safety.
For years, autonomous vehicle safety discussions have centered on a familiar question: Are autonomous vehicles safer than human drivers?
The industry has invested enormous effort trying to answer that question. Operators publish safety reports. Researchers analyze crash data. Regulators evaluate deployment programs. Insurers study emerging risk patterns.
Yet despite increasing evidence that autonomous systems may reduce certain types of crashes and injuries, public trust remains stubbornly low.
This raises an important question: what if the challenge facing autonomous vehicles is no longer only about safety performance? What if it is increasingly about how safety is evaluated, interpreted, and trusted?
The Assumption Behind Most AV Discussions
Many discussions about autonomous vehicles follow a simple logic: improved safety performance → increased trust → wider deployment.
The assumption seems reasonable. If autonomous vehicles become safer than human drivers, adoption should naturally follow. But reality appears more complicated.
The challenge may not be that safety is not improving. The challenge may be that different stakeholders evaluate safety using different evidence.
Four Stakeholders. Four Evidence Models.
Operator Evidence Model
Operators ask: are we reducing risk? They trust crash reduction, injury reduction, intervention frequency, operational performance, and fleet exposure data.
Regulatory Evidence Model
Regulators ask: how can improvement be demonstrated? They trust safety cases, hazard identification, validation, traceability, and governance processes.
Insurance Evidence Model
Insurers ask: can future losses be predicted? They trust claims frequency, claims severity, exposure data, trend stability, and loss predictability.
Public Evidence Model
The public asks: why should I trust this system? They rely on personal experience, visible incidents, media coverage, perceived control, and emotional confidence.
The Safety Evidence Gap
Consider a hypothetical announcement: autonomous vehicles reduced injury crashes by 80% compared with human drivers.
Operators may view this as compelling evidence of progress. Insurers may ask to see claims data. Regulators may ask to review the supporting safety case. The public may focus on a recent incident that received national media attention.
Everyone is discussing safety. Yet they are not discussing the same evidence.
Why This Matters
As autonomous systems mature, technical performance alone may not determine deployment success. The ability to communicate safety improvement across different stakeholder groups may become equally important.
The challenge is not necessarily that autonomous vehicles are unsafe. The challenge may be that the evidence used to demonstrate safety is not universally trusted.
This creates a gap between safety performance, safety confidence, and safety acceptance.
Key question: If autonomous vehicles eventually become demonstrably safer than human drivers, what evidence will be required for society to believe it?
Next in the series
Framework #2 explores where safety evidence begins: operational events transformed into scenario intelligence.