An Asian man driving a Lexus loaner got the now-familiar dashboard scolding: “Closed Eyes Detected. Look Forward.” His eyes were open. The infrared camera just didn’t like the geometry.
The video spread quickly. Predictable reactions followed. Some called the Japanese luxury brand racist. Lexus, founded in Japan and selling cars built around Japanese engineering, suddenly stands accused of anti-Asian bias because its software measures eyelid opening the same way most driver-monitoring systems do.
The technology is straightforward. These cameras track the visible gap between the eyelids relative to eye width. If that ratio stays below a set threshold long enough, the car assumes the driver is dozing. Many East Asian faces have a narrower palpebral fissure even when fully awake. The algorithm wasn’t trained to treat that as the default “open” state. It isn’t malice. It’s a threshold problem.
Similar videos have already circulated from Xiaomi cars in China, where drivers with smaller eyes had to deliberately widen them to silence the alerts. Lexus owner forums show the same complaints from tall drivers, glasses wearers, and anyone whose face sits at an angle the camera wasn’t optimized for. The system also struggles with sunglasses, hats, makeup, and sunlight. None of those cases produced national lectures about systemic hatred.
Calling it “racist” is the easy move. It turns a calibration issue into a morality play. The harder conversation is that these cameras are spreading because regulators want them. Europe already mandates advanced driver-distraction warnings. Japan is moving toward requiring similar systems. In the U.S., drunk-driving detection rules hang over the 2027 model year even if NHTSA keeps missing deadlines. Insurance companies love the data. Automakers add the hardware because it unlocks higher-level driver-assist features.
The result is a camera pointed at your face every time you drive, scoring whether you look “alert enough” according to software written by engineers who optimized for the faces they had in the training set. When the software misfires on a common anatomical variation, the first instinct in some circles is not “fix the threshold” but “denounce the company.
”Lexus manuals already list “eyes only slightly open” and “laughing” as conditions that can trigger false warnings. Drivers have learned to sit farther forward, ditch certain glasses, or just turn the alert off in settings. That last option still exists, for now.
The irony writes itself. A Japanese brand gets lectured by Western commentators for failing to perfectly accommodate East Asian facial features in a safety system Japan helped pioneer two decades ago.
Meanwhile, the same commentators rarely mention that the alternative is more cameras, more data collection, and less room to simply drive without a computer deciding whether your eyelids meet its standard.
The car isn’t racist. The algorithm is limited. Pretending otherwise doesn’t make the camera any smarter, and it doesn’t make the coming mandates any less intrusive.
A note on comments: Our advertisers enforce a strict zero-tolerance policy on racism, hate speech, and threats of violence. Any comment that crosses that line will be removed and the account permanently banned — no warnings. Debate hard, but keep it civil.
A note on comments: Our advertisers enforce a strict zero-tolerance policy on racism, hate speech, and threats of violence. Any comment that crosses that line will be removed and the account permanently banned — no warnings. Debate hard, but keep it civil.