Federal regulators found that self-driving AI systems were involved in almost 400 car crashes in the last year, with Teslas accounting for 70 per cent of the accidents.
The New York Times reported that Wednesday the National Highway Traffic Safety Administration shared the findings of a large-scale data collection and analysis effort to better understand the efficacy and safety of self driving systems on the road today.
According to the report, between 1 July of last year and 15 May of this year, there were 392 accidents involving self-driving technologies. In those accidents, six people were killed and five were seriously injured.
Of those 392, Tesla’s Autopilot, Full Self Driving mode, or at least a component of its self-driving tech was active in 273 of those crashes.
Self-driving systems — whether it be a nearly-full autopilot mode that allows hands-free driving or simply parallel parking assistance — become more prevalent in US vehicles with each passing year. Their growing presence on US roads prompted the regulators to examine their safety record.
“These technologies hold great promise to improve safety, but we need to understand how these vehicles are performing in real-world situations,” Dr Steven Cliff, the NHTSA's administrator, said. “This will help our investigators quickly identify potential defect trends that emerge.”
Dr Cliff spoke with reporters Wednesday about the agency’s findings, but warned not to draw too many conclusions from the numbers just yet. He noted that the data does not include information that could help further contextualise the data, like how many self-driving cars are in operation per manufacturer.
Ford, GM, and other manufacturers have been rolling out new vehicle models that incorporate self-driving technology, but it is unclear how many of those kinds of vehicles are actually on the road. Tesla has 830,000 cars in circulation with Autopilot functionality, potentially explaning why it accounts for nearly 70 per cent of the crashes recorded.
“The data may raise more questions than they answer,” Dr Cliff said.
He said the agency would continue to collect data to better understand the technology and its potential hazards.
Tesla’s self-driving technology was already facing scrutiny from the NHTSA prior to the release of the findings. The company has been the focus of a “preliminary evaluation” by the agency, which is analysing accidents involving the cars and how their self-driving tech influenced those incidents. After finding “patterns in system performance and associated driver behavior” that contributed to the accidents, the agency recommended an upgrade from a preliminary evaluation to an “engineering analysis,” which often times precedes a recall.
The agency found that in some instances Tesla Autopilot was deactivating less than one second before crashes.
"The agency's analysis of these sixteen subject first responder and road maintenance vehicle crashes indicated that Forward Collision Warnings (FCW) activated in the majority of incidents immediately prior to impact and that subsequent Automatic Emergency Braking (AEB) intervened in approximately half of the collisions. On average in these crashes, Autopilot aborted vehicle control less than one second prior to the first impact," the agency’s report reads.
