Picture a cold morning outside Shenzhen. A BYD hatchback sits in a driveway, and a stray cat has tucked itself into the flat, sheltered space beneath the front bumper. The driver climbs in, turns the car on, and nothing happens. Not a dead battery, not a software bug. The car itself decided nobody is going anywhere until whatever it detected underneath gets up and leaves.
That is the plain-language version of a patent application BYD filed with China’s National Intellectual Property Administration, published on June 12, 2026, under application number CN122200729A. It describes a computer-vision system that photographs a vehicle’s underside every time the car is switched off, compares that photo against live camera footage later, and uses the differences to work out whether something living is hiding underneath. In the most severe case, it stops the car from starting at all.
It’s a story built for a viral repost: cute, a little absurd, easy to boil down to “AI car saves cats and kids.” That reaction is fair enough. It also misses the real story, which has less to do with what BYD invented and much more to do with what a patent application actually is, why this particular filing became public in record time, and what it means that a company sitting on more than 50,000 patents just filed one more.
What The Patent Actually Describes
The system works off a simple trick: it captures a reference photo of the underbody while the car is parked and powered down, when the suspension, battery pack, and aero panels are all sitting still. When the car wakes back up, live camera footage gets compared against that reference, pixel by pixel, to build what the filing calls a difference map. Anything that hasn’t changed, meaning the fixed hardware underneath the car, gets ignored. Only the pixels that changed get run through the actual recognition step that decides whether it’s mud, a shadow, or something alive.
That’s a genuinely clever piece of cost engineering. Instead of running an AI model over an entire underbody feed continuously, which burns processing power most cars can’t spare, the system spends its compute budget only on the small patch of image that actually changed. It also means BYD isn’t proposing new hardware. The same surround-view and parking cameras many of its cars already carry can, in theory, do the looking.
The response side is where it gets more interesting. According to the filing, what the car does next depends on both the category of object it thinks it sees and how long that object lingers. A brief, low-risk detection triggers just a sound and a light. A longer stay, or a higher-risk category, adds a push notification to the owner’s phone. The most serious category, which the filing frames in a way that strongly implies a small child, restricts the vehicle from starting entirely and keeps alerting the owner’s phone until whatever is under the car leaves. The response also changes based on what the car itself is doing: a vehicle already underway just downgrades to a phone alert rather than intervening in a moving car, while a locked, parked vehicle gets the full response, including the block on starting.
A Patent Application Is Not A Product
Here is the detail that most of the viral posts skipped: CNIPA’s own listing marks CN122200729A as pending, not granted. It is an application, not an issued patent, and China’s patent office publishes applications long before anyone decides whether they hold up. CarNewsChina, which covered the filing directly, put it plainly: publication of a patent application “does not indicate production deployment.” No vehicle program, no launch window, no commercialization plan appeared anywhere in the document.
That distinction matters more with BYD than with almost any other automaker on earth. An analysis of the company’s patent portfolio published in October 2025 put its worldwide filings at more than 50,000, with over 30,000 already granted. The World Intellectual Property Organization has separately found that roughly 82 percent of BYD’s patents are only active inside China, with just 7 percent active in the United States. BYD does not sell passenger vehicles in the U.S. market at all right now, tariffs have seen to that, so a huge share of what the company patents is, by design, aimed at a domestic audience and a domestic strategy.
Why It Went Public So Fast
Here is the first wait, really moment. Under Chinese patent law, an invention application is normally published around 18 months after its earliest filing date, unless the applicant specifically asks for it to come out sooner. BYD filed this application on January 30, 2026. CNIPA published it on June 12, 2026, about four and a half months later. That is a fraction of the standard timeline, and it does not happen by accident. Someone at BYD requested early publication, which reads a lot less like a company protecting a trade secret and a lot more like a company that wanted the attention now, while the EV world is watching every move it makes.
The Part Everyone Skipped: A Car That Can Refuse To Drive
Any camera system built to tell a sleeping cat apart from a wet leaf pile or a hard shadow is going to misfire sometimes. The patent’s own background section admits as much, describing existing underbody detection attempts as suffering from “insufficient all-weather reliability,” plagued by low light, road debris, and inconsistent ground surfaces. BYD’s fix is clever, but cleverness doesn’t erase false positives, it just reduces them.
That matters because this isn’t a system that just annoys you with a chime. In its most aggressive form, it immobilizes the vehicle. Picture the inverse of the cute scenario: a driver late for work, a car that won’t start because its cameras mistook a shadow or a shopping bag for a living thing, and no clear way to override it in the moment. Once an automaker builds a system that claims to decide for the driver, the automaker also inherits the blame when that system decides wrong. That is a lesson the industry has already learned the hard way with driver-assistance features that overpromised on what they could reliably see.
The Blind Spot No Regulator Has Touched
American regulators have been moving in a related direction, just not this direction. NHTSA finalized a rule requiring rear seat belt warning systems on new vehicles starting September 1, 2027, and the agency has spent years pushing toward technology that can sense whether a child has been left inside a vehicle, a response to hundreds of hot-car deaths. That’s the same agency we’ve watched push automakers on other overlooked failure points, from door latches to an 807,000-vehicle airbag investigation into older Honda Odyssey minivans. None of that regulatory attention reaches underneath a parked car. What BYD’s patent targets, a small child or animal sheltering beneath a vehicle that’s about to move, sits in a genuine, decades-old blind spot that mechanics have warned about for generations without any binding rule anywhere addressing it.
Here’s the second wait, really moment. The classic version of this hazard is a cat crawling up into a warm engine bay in winter, which is why mechanics have spent decades telling owners to bang on the hood before starting a cold car. BYD’s own vehicles largely sidestep that specific danger since electric motors don’t throw off the same lingering heat, but BYD’s Blade Battery architecture, which integrates the battery pack into the floor structure itself, creates an enormous, flat, sheltered underside running nearly the length of the car. Swapping one kind of attractive hiding spot for another isn’t a flaw in the engineering. It’s just a reason the underbody problem hasn’t gone away with electrification, it’s changed shape.
Why A Filing Cabinet In Shenzhen Is Now Automotive News
Zoom out and this patent fits a pattern The Auto Wire has been tracking all year. Chinese automakers are winning the current phase of the EV war on cost and feature velocity as much as on range or horsepower, a shift we’ve followed as the platforms behind familiar-sounding brands keep tracing back to China, and as legacy players like Volkswagen suddenly find their own bestsellers outsold at home. Sensing and software claims, real or theoretical, have become part of that competition. A viral patent costs a fraction of what a production feature does, and it generates the same headlines.
It’s the same instinct behind Hyundai and Kia’s push into self-disinfecting cabin UV lights, a feature announcement that, on inspection, said more about staffing and cleaning costs than about germs. Automakers have realized that a specific, slightly strange, easy-to-visualize safety claim travels further on social media than a horsepower figure ever will, whether or not it ships.
What To Actually Remember Here
None of this means the idea is bad. Underbody sheltering by kids and pets is a real, under-addressed hazard, and using cameras a car already has, instead of bolting on new hardware, is a smart way to attack it cheaply. If BYD or anyone else actually engineers this into a reliable production system, it deserves credit for closing a blind spot the rest of the industry has ignored for decades.
But that’s a separate question from the one this week’s headlines actually answered. BYD didn’t build a car that checks under itself for kids and cats. It filed the paperwork for one, asked China’s patent office to publish it in a fraction of the usual time, and let the internet do the rest. Somewhere between a patent examiner’s desk in Beijing and a viral repost, an idea became a feature before it became a product. That gap, not the sensor, is the story worth remembering.