Gasgoo Munich-Promises of hands-free driving versus liability clauses: which truly represents L2 driver assistance?
In late May, BYD Chairman Wang Chuanfu announced at a smart strategy launch that users equipped with the DiPilot A and B systems would be fully compensated for direct economic losses if an at-fault accident occurred while using the city navigation function compliantly.
Less than a month later, Huawei's Yinwang followed suit, extending coverage to urban NCA, highway NCA, and full-scenario parking. With two industry leaders stepping up to the plate, the long-standing question of who pays when driver-assist systems fail has been thrust into the spotlight.
Public opinion is divided. Some applaud the move, seeing it as automakers finally footing the bill for their technology; others dismiss it as merely a rebranded "smart driving insurance."
More telling is the silence from the rest of an industry known for copycatting. XPeng, NIO, and Li Auto—key players in intelligent driving—have largely stuck to paid insurance models, declining to offer similar direct guarantees.
This clash between bold launch promises and strict liability clauses, where two leaders pay up while peers stay quiet, signals a shift far deeper than a mere marketing stunt.
Cao Guangping, a partner at Cheshang Consulting, argues that the split—where a few firms underwrite risk while most do not—reflects the current reality of China's intelligent driving technology, market, marketing, service, and legal landscape.
Image source: BYD
The L2 Liability Black Hole
For years, China's auto sector has waged a fierce battle over driver assistance. Ads show steering wheels turning automatically and seamless lane changes, reinforcing the perception of "smart driving" while drivers work or relax.
Yet open any owner's manual, and the warnings are stark: "L2 driver assistance requires full driver responsibility," "The system assists, it does not replace; hands must remain on the wheel."
Automakers sell the dream of "smart driving" but retreat to "driver assistance" when things go wrong. They reap the premiums and traffic, while consumers shoulder all the risk.
This gap between hype and reality has led to frequent crashes. In the first half of 2025 alone, 137 accidents involving driver-assist systems occurred nationwide—a 43% year-on-year jump.
In March 2025, a Xiaomi SU7 failed to recognize a construction detour on the highway, exiting at the last moment and causing a collision. Then in September, Xiaomi recalled over 110,000 SU7 units because some batches struggled to identify extreme scenarios during L2 highway navigation.
Incidents keep happening: systems missing stopped construction vehicles, failing to sense static obstacles in rain, or making erratic lane changes. Each incident forces the public to ask: Can L2 really be trusted?
Legally, the lines are clear. In early 2026, the Supreme People's Court ruled that in-car driver-assist systems cannot replace the driver as the primary operator. The person activating the system remains responsible for safety. In one dangerous driving case, a driver who activated the system while drunk and moved to the passenger seat was still convicted.
The legal bottom line is drawn. The problem is that consumers hear a different narrative. Some automakers downplay the "assist" aspect while hyping "automation" to the skies.
Despite the Ministry of Industry and Information Technology banning exaggerated claims, liability rules remain traditional. When systems misjudge, the driver pays, and automakers rarely face consequences for overpromising. This imbalance is the black hole at the heart of L2 liability.
Cao Guangping notes that the legal reality for users boils down to four points: for L2, the driver is always liable; for L3, the driver is liable when manually driving; for L3, the driver is liable when switching to machine mode; and for L3, the driver is liable when the machine requests intervention.
Meanwhile, usage is surging. In the first four months of 2026, the installation rate of L2 systems in new energy passenger vehicles hit 58.2%, with L2+ reaching 29.2%. HarmonySmart Mobility saw a 98% user activity rate for smart driving throughout 2025.
As more people use these features, two questions loom large: How is safety guaranteed? Who pays when things go wrong? These are no longer niche concerns but realities the industry must address.
How BYD Covers, How Huawei Follows
In late May, at a "Dare to Act" themed event, BYD's Wang Chuanfu reiterated the promise: for users of DiPilot A and B, any at-fault accident during compliant city navigation use would be fully compensated by BYD, with no cap and without tapping the owner's commercial insurance.
This wasn't BYD's first such promise; in July 2025, it introduced a safety guarantee for smart parking.
But the guarantee isn't unconditional. Vehicles with the DiPilot C system require a 12,000 yuan upgrade to qualify—up from the previous 9,900 yuan option. Meanwhile, BYD made the DiPilot B LiDAR system available across its lineup for 12,000 yuan.
Image source: BYD
BYD's confidence rests on over a decade of tech accumulation and full-stack autonomy. Wang unveiled the Xuanji A3, China's first 4nm automotive-grade smart driving chip, featuring a 16-core CPU, 273GB/s bandwidth, and over 2,100 TOPS of combined computing power with three units.
BYD boasts a fleet of over 3.15 million intelligent driving vehicles, generating 200 million kilometers of data daily. With a 5,000-person R&D team, its scale has driven down per-vehicle costs. Its vehicle intelligence strategy integrates architecture, powertrain, and chassis, giving it control over the entire chain from hardware to algorithms.
Huawei is moving even faster. On June 22, Yinwang announced new ADS high-level package guarantees, covering urban NCA, highway NCA, and full-scenario parking.
The coverage mirrors BYD's: if an accident occurs during normal use, causing damage to the vehicle, third parties, or injury, the company bears the cost.
However, Huawei's plan has nuances. Coverage duration depends on purchase timing: users who bought the high-level package before July 1 get one year of coverage; those buying after the price hike get three years.
Huawei attributed the price adjustment to rising hardware costs. Spot prices for high-end automotive DDR5 memory have surged over 300%, as demand for high-compute AI chips squeezes automotive-grade capacity.
The buyout price for the ADS Max package reverted from 32,000 yuan to the standard 36,000 yuan, with the post-subsidy price rising from 12,000 yuan to 15,000 yuan.
Huawei's confidence is also data-driven. By May 2026, its Qiankun ADS system had logged 114.7 billion kilometers of assisted driving, with a 95.1% monthly active user rate. Yinwang stated the guarantee is based on confidence derived from massive real-world data.
Image source: Huawei Qiankun
Despite differences, both moves point in one direction: automakers are starting to put real money behind the risks of their driver-assist systems.
Previously, the industry standard was "smart driving insurance," handled by insurers. BYD and Huawei are effectively shifting from technology providers to risk bearers.
But a distinction remains: these guarantees cover property and injury damages. Criminal liability and traffic points still fall to the driver, as the Supreme Court's rulings have made clear. In short, automakers are underwriting the financial risk, not the legal one.
Cao argues that while these guarantees mark technical progress, they are also a defensive response to competitive pressure from Tesla's incoming FSD. The fact that most firms haven't followed suit reflects the reality of the industry's service capabilities.
In his view, automakers can only reasonably underwrite risk when the vehicle is in machine-driving mode, and even then, only for their own vehicle's liability—not for all parties involved. Essentially, it's half service, half technology, and any disputes will ultimately be settled in court.
Why Are Other Automakers Holding Back?
BYD and Huawei have made the first move, but the reaction from others has been surprisingly cold.
In April 2025, XPeng launched a Smart Driving Peace of Mind service for 239 yuan a year, offering up to 1 million yuan in coverage for NGP and parking. It even covers accidents within 5 seconds of the system disengaging. Yet, fundamentally, this remains an insurance product, affecting the owner's claims record and future premiums.
NIO and Li Auto also rely on paid subscriptions for driver-assist coverage, featuring payout caps and requiring users to claim through their personal insurance first.
Why haven't XPeng, NIO, and Li Auto—recognized leaders in this field—followed suit with direct guarantees? The reasons can be broken down into three layers.
First, the math doesn't work. Underwriting risk requires a massive support infrastructure for customer service, claims assessment, repairs, and legal control. For brands without sufficient scale or penetration, the numbers simply don't add up.
BYD can afford it because it has over 3.15 million vehicles to spread the risk. Without that scale, jumping in could drag a company into a financial quagmire.
Image source: XPeng
Second, the technology isn't ready. BYD's guarantee is built on full-stack self-development, with its Xuanji 2.0 architecture and mass-produced A3 chip. Huawei has its Qiankun ADS validated by billions of kilometers.
Most automakers, however, rely on supplier solutions and have limited control over system boundaries and failure scenarios. Naturally, they hesitate to guarantee a system they don't fully control.
Third, the legal coverage is incomplete. Even with financial compensation, criminal liability and points stay with the driver. The Supreme Court's rulings are non-negotiable.
For automakers, this is a commercial promise, not a legal shield. It fails to fully resolve the driver's liability anxiety and could spark new disputes over policy boundaries. Many prefer the status quo—letting users buy insurance—rather than taking on uncertain reputational risk.
Past price and configuration wars were easy to follow: add a radar, boost computing power. But underwriting risk is a promise of unlimited liability. The dimension of competition has shifted from hardware to responsibility, and the cost of following is entirely different.
There's also the broader context: in 2026, costs for automotive chips and battery materials kept rising, squeezing profit margins. Under pressure to survive, most automakers are focused on preserving cash, not adding uncertain financial burdens.
Cao concludes that the key to whether tech can enable such guarantees lies in whether drivers can truly disengage their brain, eyes, hands, and feet across L2 and L3 modes—and what the real takeover rate is.
Conclusion
BYD and Huawei have started the conversation, pushing the question of liability to the forefront. But two players are far from enough.
With L2+ installation rates hitting 29.2% in the first four months of 2026 and millions using these features daily, the industry needs more than isolated promises. It needs a replicable mechanism for shared responsibility.
The industry's "involution" should focus on substance: quality, responsibility, and consumer safety. When others follow depends on when cost, technology, and legal barriers can be overcome.
Until then, underwriting L2 liability remains a game for the few.
