On 10 September, nine Chinese ministries published a five-year plan that sets a date for self-driving cars. By 2030, vehicles with autonomous driving functions are to reach what the document calls scale application, with high automation achieved on motorways, urban expressways and some city streets.
Ten weeks earlier, on 26 June, the American government withdrew its only proposed programme for overseeing automated vehicles.
These two facts are usually filed under the same headline about a race. They are not a race. They are two countries using instruments so different that the comparison barely holds.
What the Chinese document actually says
The plan is the 15th Five-Year Plan for the Intelligent Connected New Energy Vehicle Industry, document number 2026 No. 305, signed 9 September and issued jointly by nine bodies including the Ministry of Industry and Information Technology, the National Development and Reform Commission, the Ministry of Public Security and the Ministry of Transport.
Its 2030 targets are specific. New energy vehicles are to reach 70 percent of domestic passenger car sales and 40 percent of commercial vehicle sales. Average passenger car fuel consumption falls to 3.3 litres per 100 kilometres. Labour productivity rises 15 percent against 2025. Several Chinese firms are to rank in the global top ten by sales.
And then a line that is easy to skim past: vehicles fitted with autonomous driving systems must have safety performance that substantially exceeds human drivers.
That is a state-defined safety threshold for machine driving, written into a national plan with a date attached.
The NEV target is already met. The autonomy target is the real one.
Here is the first thing that reframes the document. That 70 percent NEV figure for passenger cars sounds ambitious until you check where China already is.
Industry figures for August 2026 put new energy vehicles at 65.7 percent of domestic sales, with passenger cars specifically at 68 percent. The five-year target is roughly two points above what the market did last month.
So the electrification number is not a stretch goal. It is a floor, written to be met. The binding commitment in this plan is the autonomy line, and that is the one worth watching.
China already has the mandatory standard. That is the part America lacks entirely.
Targets are cheap. What makes this plan different from a press release is the regulatory machinery that arrived ahead of it.
In June 2026 China published GB 47955-2026, a mandatory national standard for combined driving assistance systems, in force 1 January 2027. In July 2026 it published GB 44721-2026, a mandatory national standard for Level 3 and Level 4 automated driving systems, in force 1 July 2027. Note the absence of a slash in those numbers: in the Chinese system, GB means mandatory and GB/T means recommended.
GB 44721 sets a performance floor in law: the automated driving system must at minimum match a competent and attentive human driver performing the same dynamic driving task. It requires enterprise lifecycle safety management, human-machine interface rules, and multi-level verification through simulation, track and road testing with third-party confirmation.
Now the American position, in the government’s own words. From NHTSA’s 2025 proposal: “The FMVSS do not currently set performance standards specifically for ADS,” and compliant vehicles “can generally be equipped with ADS technologies without NHTSA approval.”
There is no American equivalent of GB 44721. There is no federal performance standard for automated driving at all. You can fit a self-driving system to a compliant car and no one has to approve it.
The programme that died ten weeks before the Chinese plan
The United States did try. In January 2025, NHTSA proposed AV STEP, a voluntary national programme for oversight and transparency of ADS-equipped vehicles.
On 26 June 2026 it was withdrawn. The agency gave three reasons, and the middle one is the most telling: industry said a voluntary programme was too burdensome to join, while safety advocates said a voluntary programme was not a sufficiently comprehensive regulatory tool.
A proposal that was simultaneously too much and not enough died of being voluntary. What survived, NHTSA said, was the work of expanding and improving exemption processes.
Which brings us to the instrument America actually uses.
Two thousand five hundred, two years at a time
A purpose-built driverless vehicle with no steering wheel cannot comply with federal motor vehicle safety standards written around a human driver. The only route to market is an exemption under 49 U.S.C. 30113.
That statute caps such exemptions at 2,500 vehicles sold in any 12-month period, and limits them to two years at a time. Congress set that number long before anyone was building a car without controls.
The scale comparison is the part that should stop you. China sold 1.643 million new energy vehicles in the single month of August 2026. NHTSA, by its own account, exempted 347 imported ADS-equipped vehicles across 295 projects in 31 states over the nine years from 2016 through 2024.
One country regulates by five-year plan. The other regulates by permission slip, one applicant at a time, with a ceiling Congress wrote during the Johnson administration.
The rule that makes the whole race question moot
And now the fact that quietly dissolves most of the commentary about who is winning.
In January 2025 the Commerce Department’s Bureau of Industry and Security published its connected vehicles rule, effective 17 March 2025. It prohibits importing or selling in the United States completed connected vehicles incorporating software that supports vehicle connectivity systems or automated driving systems with a Chinese or Russian nexus. Automated driving system software is covered from model year 2027. Connectivity hardware follows at model year 2030.
The rule also bars manufacturers subject to Chinese jurisdiction from offering commercial services in the United States using vehicles that incorporate an automated driving system. Not just selling the car. Running the service.
So whatever China deploys at scale in 2030 is, by American law, already excluded from the American market. The same rule is why Waymo has been stockpiling Chinese-built robotaxis in Arizona. The two systems are not competing for the same customers. They have been legally separated, and the separation takes effect at a model year, not a calendar date.
What China means by deployed, and what it has actually deployed
It is worth being precise about how far along China actually is, because the gap between the plan and the fleet is instructive.
On 15 December 2025, MIIT issued China’s first conditional licences for Level 3 production and road use. Not to an industry. To two models: a Changan battery-electric sedan and an Arcfox battery-electric sedan.
Their permitted operating domains are narrow to the point of being addresses. The Changan is approved for single-lane automated driving in congested traffic on three named Chongqing road segments. The Arcfox on three named Beijing expressways. Business users first, individual owners phased in later.
Two cars, six stretches of road. Compare that with Waymo being cleared for paid driverless rides across 18 California counties in a single regulatory decision. That is China’s deployed Level 3 fleet as of the start of this year. The country is not ahead on cars on the road. It is ahead on having a legal pathway and a date.
The line about standards that deserves more attention than the autonomy target
Tucked into the same paragraph as the fuel consumption figures is an objective with a longer reach than anything else in the plan: that China’s voice in setting international standards and regulations be further raised by 2030.
Consider what that means structurally. Europe writes vehicle automation rules through UNECE, where the published text of Regulation 157 on automated lane keeping still reads 60 km/h and category M1, with a higher-speed amendment applying only in countries that choose to adopt it. America self-certifies against standards that say nothing about automated driving. China writes mandatory GB standards and has noted publicly that its new assistance standard goes beyond the UN equivalent on operating range, driver-state detection and user notification.
Three incompatible rulebooks. One of them is attached to the world’s largest vehicle market and comes with a written intention to export itself.
What to remember
Forget the target percentages. Remember the difference in instruments, because that is what will still be true in 2030.
China has published a date, a mandatory safety standard that takes effect before that date, and a definition of how safe a robot driver has to be. America has an exemption statute, a 2,500-unit annual ceiling, a two-year clock, and a withdrawn proposal.
Whoever writes down what safe enough means gets to define it for everybody who wants to sell there. Right now only one of these countries has written it down.
Should the US move faster on self-driving rules, or is the slow approach the safer one? Share your thoughts in the comments.