The rapid deployment of autonomous driving features on Chinese cars has sparked alarm among regulators in Beijing, who have made the industry tap on the brakes while they assess questions over safety and liability.
Despite an unclear legal framework for new assisted-driving technologies, nearly one in five new cars sold in China is now equipped with high-level autonomous functions.
Beijing officials, caught on the back foot, are expected to slow the rollout as they develop a regulatory framework for the new suite of technologies, where China is rapidly becoming a world leader.
“The cat is already out of the bag; they’re not going to try to put it back in,” said Tu Le, founder of the Sino Auto Insights consultancy. “But what regulators may do is limit use cases for the foreseeable future, until they have a much better understanding of the implications.”
In March, three people were killed in an accident involving a Xiaomi SU7 electric sedan with semi-autonomous capabilities. This fuelled an intense debate online over safety and sparked a warning from Beijing on overzealous marketing of unproven self-driving technologies.
This week, China’s Ministry of Industry and Information Technology proposed the development of new safety requirements for driver assistance systems, a forerunner to driverless cars. New standards aim to reduce accidents and guide technology development, while reducing research costs for companies, according to a notice published online.
While there remains high-level policy support for China’s electric vehicle industry, the central government has not given a clear signal on the timing or scope of broader, nationwide regulations for fully autonomous vehicles, mostly deferring to local governments to oversee pilot projects.
HSBC analyst Yuqian Ding said Beijing had sought industry “self-discipline” as new regulations are developed. But in the eyes of auto executives in China, she said there was “no longer” a question over whether the technology is needed amid cut-throat competition. “Everyone agrees: ‘If I don’t do this, I’m not going to survive’,” she said.
China’s insurance and transport regulators have traditionally looked to best practices overseas to help form their approach on key rules around liability, such as who bears responsibility for accidents and damage, and how to price insurance.
Some content could not load. Check your internet connection or browser settings.
However, with China leading in driverless technology and Beijing insisting on the need to align with its strict data-security controls, the country may be forced this time to develop industry-wide regulations ahead of other jurisdictions.
The development of a new legal framework faces further complications from the likelihood of a years-long transition, when there will be a mix of human drivers and fleets of robot cars transporting people and goods.
Sam Radwan, principal at Enhance International, a consultancy advising Chinese insurers, predicted that a final insurance system for increasingly smart cars could still be five or 10 years away.
There was a fundamental problem, he said, in deciding how to price risk and allocate responsibility when a vehicle’s operating system was frequently downloading software updates. They potentially changed a vehicle’s capabilities and the level of attention and interaction required by a driver.
However, “statistically, driverless cars are going to crash less than cars driven by people, so premiums for the auto industry are likely to, over time, be reduced”, said Radwan. “But it’s a messy period getting there.”
The patchwork of current rules on liability implies that, with Level 3 vehicles, which can manage complex scenarios but require drivers to be ready to take control upon a system prompt, responsibility is shared among drivers, insurers and automakers.
At Level 4 — full self-driving in specific conditions in vehicles such as robotaxis — accident responsibility is expected to sit with fleet operators. Yet there are still questions over whether hardware and software suppliers should also be held liable.
In an industry where the US and China are competing for supremacy, regulatory questions are also emerging in America, such as whether Cybercabs developed by Elon Musk’s Tesla would be allowed to drive on American roads without pedals or a steering wheel.
Lei Jun, founder of smartphone giant-turned-carmaker Xiaomi, is among those who have urged the government to adopt a new nationwide safety testing and verification system for autonomous vehicles, as well as a compulsory insurance system for the new breed of cars.
Paul Gong, who leads China automotive analysis at UBS, said there might still be merit in the lack of clarity in China at this stage of the industry’s development.
“Excessive regulation can hinder technological development, as we’ve seen in Europe and the US. It’s more important to let engineers drive progress rather than letting lawyers manage it.”
Some content could not load. Check your internet connection or browser settings.
Autonomous driving pilot zones have been approved in about 20 Chinese cities, including large-scale robotaxi tests in Beijing, Shanghai, Shenzhen, Chongqing and Wuhan.
Goldman Sachs has forecast China’s robotaxi fleet would pass 500,000 cars over the next five years, while it expects the Chinese robotaxi market to be worth $47bn by 2035, up from just $54mn this year.
James Peng, chief executive of Chinese robotaxi start-up Pony.ai, said while there appeared to be a “clear road map” for expanding the operational regions for the L4 vehicles, regulations over L3 were a “grey area” and “even the product definition is still not very clear”.
Ya-Qin Zhang, who chairs the Apollo alliance, an open autonomous driving platform led by search group Baidu, is more sanguine about the regulatory outlook. He believes the current approach seeks to “balance safety and innovation”.
Zhang, who also leads the Institute for AI Industry Research at Tsinghua University, described 2024 as “explosive” for the development of true driverless technology in China, thanks to the coupling of mass data collection from the pilot zones and new powerful AI applications that allowed faster training for driverless cars and helped them to reason and react in a more human way.
He expects that by 2030, about 10 per cent of China’s new cars will have the capability to be operated without a driver. And by that time, the regulatory framework will be ready.
“We need cars to learn how humans drive,” Zhang said. “The [pilot zones] collect a lot of human driving behaviour.”
Additional reporting by Kana Inagaki in London