New data suggests that autonomous vehicles are 90% safer than their human-driven equivalents
Join other savvy professionals just like you at CIHT. We are committed to fulfilling your professional development needs throughout your career
By Andrew Mylius
If you live or work in London or Manchester, you have probably seen a ‘robotaxi’. You might even have ridden in one if you are an Uber user. These autonomous vehicles (AVs) have been collecting data on the streets of the capital for more than a year, while trials have just begun in Manchester.
Though the taxis are self-driving, they currently have human ‘safety drivers’ standing by to intervene if required, and there are several licensing and regulatory hurdles to clear before wider implementation happens. Among the hurdles, the biggest is proving that AVs are safe – a look across the Atlantic suggests they are.
In December 2025 the US Autonomous Vehicle Industry Association (AVIA) published its first Robotaxi Report, which detailed explosive growth in America’s self-driving taxi market from $450m in 2024 to a projected $10.39bn by 2030 and $400bn by 2035. The report also includes a study into the safety performance of Waymo self-driving taxis, particularly in San Francisco.
AVIA noted that Swiss Re, one of the world’s leading reinsurers, had analysed insurance claims clocked up over the course of 25.3 million fully autonomous Waymo miles. Swiss Re used the claims as a proxy for at-fault collisions, comparing them with a wider dataset of claims representing more than 200 billion miles of human driving.
The comparison revealed that, mile for mile, Waymo’s vehicles were involved in 88% fewer collisions resulting in property damage claims and 92% fewer collisions resulting in bodily injury claims compared to human-driven vehicles.
These safety advantages even held up against the latest human driven vehicles equipped with advanced driver assistance systems, with Waymo showing an 86% reduction in property damage claims and a 90% reduction in injury claims.
Appointed by AVIA to examine the economic consequences, AECOM calculated that Waymo drove 29.88 million miles in San Francisco between March 2022 and June 2025. Compared to human-driven vehicles, the amount of collisions avoided equated to an estimated $71m in safety-related economic savings.
Dr Nick Reed, a leading mobility advisor and consultant, notes that “the early picture painted by safety performance data is strikingly positive”. However, strong safety performance can be overshadowed by “unusual incidents”, such as an AV’s failure to register the approach of an emergency vehicle and move out of its way.
“It’s difficult to build public confidence in automated vehicles,” says Reed. “Because if there are high-profile incidents where they do unexpected things, that’s what sticks in people’s mind – no matter how impressive the safety statistics might be.” Reed also notes that road safety standards are significantly higher in the UK compared to Waymo’s US markets, so care is needed when estimating how safety improvements might translate to British roads.
At present, AV taxi providers are pursuing different sensing technologies, operating concepts and system architectures, with each having to prove they can meet the safety outcomes of the markets they want to operate in. Regulators are increasingly focused on whether all providers can demonstrate equivalent standards of safe operation.
One area where divergence could prove to be problematic is an AV’s interaction with emergency services, “It’s vital that firefighters, police and paramedics can do their job of saving lives,” says Reed, “securing evidence and restoring traffic flow without delay. Standardisation of protocols and collaboration between technology developers and the emergency services will help to ensure that any encounters involving automated vehicles can be resolved successfully.”
Artificial intelligence and self-learning will be highly influential in accelerating improvements in safety, and the information-processing capabilities of AVs could mean they are held to higher standards of safety than humans, simply because they are capable of achieving them. Reed points to the aviation industry as a model worth emulating: when safety incidents occur, learning is shared and improvements arise from collective experience.
• Read CIHT’s response to the consultation on the future safety of automated vehicles here
• Take the CIHT course on AI for the highways and transportation sector here
Join other savvy professionals just like you at CIHT. We are committed to fulfilling your professional development needs throughout your career
{{item.AuthorName}} {{item.AuthorName}} says on {{item.DateFormattedString}}: