Futuristic self-driving cars travelling along California roads have needed plenty of old-fashioned human intervention to stay safe, according to test reports.
California's Department of Motor Vehicles released reports filed by seven companies given permission by the agency to test prototype vehicles in public which gave instances in which a driver had to take over because of technology problems or other safety concerns.
The reports show wildly different levels of success since on-road testing started in September 2014.
Experts in the technology said Google, whose cars drove the most by far, performed relatively well, but also cautioned that the testing typically happened during good weather.
Other companies reported frequent instances in which the person required to be in the front seat just in case had to grab the wheel.
Nissan, for example, tested just 1,485 miles in public, but reported 106 cases where the driver had to take control. The Japanese car maker has said it plans to have "commercially viable autonomous drive vehicles" by 2020.
Google said its cars needed human help 341 times over 424,000 miles - the equivalent of about 10 times a year, given the 12,000 miles the average US vehicle travels annually.
In 11 of the 341 instances, Google said its cars would have been involved in a crash.
Chris Urmson, head of the company's self-driving car project, said while the results were encouraging they also showed the technology had yet to reach his goal of not needing someone behind the wheel.
"There's none where it was like, 'Holy cow, we just avoided a big wreck'," he said.
"We're seeing lots of improvement. But it's not quite ready yet. That's exactly why we test our vehicles with a steering wheel and pedals."
The California Department of Motor Vehicles, which is writing new regulations for the technology, said it was still reviewing in the reports.
Google reported 272 cases in which the cars' software or onboard sensors failed. Though Google did not release detailed scenarios, the problems included issues with the self-driving cars seeing traffic lights, yielding to pedestrians or committing traffic offences.
There were also cases where intervention was needed because other drivers were reckless and several dozen instances of an "unwanted manoeuvre" by Google's car.
Bryant Walker Smith, a professor at the University of South Carolina who closely follows self-driving car developments, said Google's rate of potential collisions was "not terribly high, but certainly not trivial".
He said it remained difficult to gauge how self-driving cars compared with accident rates among human drivers, since even the best data under-reported minor collisions that authorities were never told about.
While Google's problem rate was "impressively low," a trained safety driver should remain in the front seat, said Raj Rajkumar, an engineering professor at Carnegie Mellon University who specialises in self-driving cars.
According to data in Google's report, a driver typically took control within one second of the car asking for help.
Drivers at other companies often reacted quickly as well, according to their reports, though Volkswagen Group of America reported that in one case, it was more than 12 minutes before the person took control.
John Simpson, a frequent critic of Google who focuses on privacy issues for the non-profit group Consumer Watchdog, said the company's report "underscores the need for a driver behind the steering wheel capable of taking control of the robot car".
Google has argued to California regulators that once the company concludes the cars are ready for the public to use, they should not need a steering wheel or pedals because human intervention would actually make them less safe.
Google released its report before the agency posted reports from other companies in what Google described as an effort to be transparent about its safety record. The company had lobbied against having to report disengagements in the first place, saying the data could be misinterpreted.
The other companies testing self-driving cars on California streets are Tesla Motors, Mercedes-Benz and parts suppliers Bosch and Delphi.
Google's testing mostly involves driving around the company's Silicon Valley headquarters or the streets of Austin, Texas.
The company's rate of human intervention has improved in recent months, according to its data, but Mr Urmson warned that the rate might again rise as Google subjected the cars to more challenging environments and weather conditions.
Google said its cars would have been responsible in eight of the 11 avoided accidents, according to computer modelling the company performed later. In two other cases, its cars would have hit a traffic cone.
Google cars have been involved in nine collisions since September 2014. In each case, the other car was responsible, according to an analysis by researchers at Virginia Tech University.