There is a camera in the Volvo EX90 that points directly at your face.
It's not the camera on the outside of the car. It's not the one that monitors your blind spots or reads lane markings. It's inside the cabin, positioned where you can't easily see it, and it runs continuously while you're driving. Volvo calls the system its "Driver Understanding System." Its job is not to record the road. Its job is to understand you — your eye movements, your head position, your alertness level, and increasingly, your emotional state — and to make decisions about your driving experience based on what it sees.
When Volvo announced this technology, a significant number of EX90 owners went online to ask where exactly the camera was. Several said they couldn't find it. Which is, depending on how you feel about being observed, either reassuring or deeply uncomfortable.
Here's the thing: Volvo isn't unusual anymore. <cite index="16-1">In 2026, many mainstream models ship with camera-based driver monitoring systems. Notable examples include the 2026 Mercedes-Benz CLA, which uses an interior camera above the centre screen, Subaru models with EyeSight and DriverFocus, Nissan with ProPILOT Assist 2.0, and BMW's Neue Klasse cars with an infrared camera in the rearview mirror.</cite> Since August 2024, when the EU General Safety Regulations made driver monitoring systems legally required on all newly designed cars, the camera watching your face while you drive has moved from premium feature to regulatory mandate.
And that camera is learning things about you that nobody has ever had access to before.
What the Camera Already Knows
The first generation of driver monitoring systems was relatively modest in scope. They watched for drooping eyelids, a forward-falling head, the specific micro-patterns of eye closure that signal fatigue. They issued a warning when the pattern matched drowsiness, and they were considered an impressive safety feature.
That was the starting point. The current generation does considerably more.
<cite index="9-1">Modern systems can accurately detect fatigue, distraction, fear, and neutrality in real time — the critical states that have a direct impact on driving safety. Unlike traditional systems that focus solely on fatigue, newer frameworks detect emotion-driven driver behaviours across a much wider spectrum.</cite>
The academic research behind production implementations is more granular still. <cite index="10-1">Studies have demonstrated a link between specific negative emotions and impaired driving performance. Sadness has been shown to increase location error rates, while anger slows drivers' ability to identify road elements. Emotions such as anger, hostility, and nervousness are strongly associated with aggressive driving behaviours. Drivers experiencing sadness, anger, or agitation were nearly ten times more likely to be involved in an accident.</cite>
Ten times. The research establishing that connection is not new or contested — it has been consistent across multiple independent studies. And now the automotive industry has the technology to detect those emotional states in real time, from a camera smaller than your thumbnail, at 98 percent accuracy.
<cite index="9-2">The proposed systems achieve an overall accuracy of 98.6 percent, with an F1 score of 0.979 across four emotional states — enabling the precise recognition of driving-relevant emotions and making them a practical, high-performing solution for real-world in-car driver monitoring.</cite>
Ninety-eight point six percent. To put that in context: a moderately attentive human being, attempting to assess another person's emotional state from their face, manages somewhere between 60 and 80 percent accuracy under good conditions. The camera in your rearview mirror is outperforming human emotional perception by a meaningful margin.
The Market Growing Around Your Face
This technology is not a research project. It is an industry.
<cite index="9-3">The global market for driver monitoring systems was estimated at USD 3.03 billion in 2024 and is expected to grow at a compound annual growth rate of 11.7 percent between 2025 and 2033, reaching USD 8.10 billion.</cite>
Three billion dollars, growing toward eight. Those numbers reflect a regulatory environment — Euro NCAP 2026 requires in-cabin monitoring compliance for a top safety rating — and a genuine industry consensus that emotional state detection in vehicles is both technically feasible and commercially viable. The question has moved from "can this be built?" to "how should it be deployed?" — and that second question has a much messier set of answers.
<cite index="17-1">Premium manufacturers have diverged in their approaches. Volvo leads with its dual-camera Driver Understanding System that reportedly exceeds compliance targets. BMW combines AI and infrared vision for what it calls a premium, human-centric experience. Mass-market manufacturers like Stellantis, Hyundai, and Toyota emphasise cost-effective, scalable solutions. Tesla pursues a distinct software-driven, camera-centric monitoring strategy leveraging fleet-scale data.</cite>
Different architectures, different data strategies, and — critically — different answers to the question of what happens to the information the camera collects.
Where the Data Goes
This is the part of the story that requires the most careful attention, because the answers vary dramatically by manufacturer and are not always easy to find in the documentation a buyer receives at the dealership.
<cite index="16-2">Some manufacturers explicitly told Consumer Reports their systems keep in-cabin data local. BMW, Ford, and GM told Consumer Reports they do not transmit video or in-cabin data off the vehicle. Subaru told Consumer Reports its DriverFocus uses facial recognition but does not record information.</cite>
That is a reasonable position, and for buyers those commitments matter. A driver monitoring system that processes data locally, makes a real-time safety assessment, and discards the input without transmitting or storing it is meaningfully different from one that retains the data and makes it available for other uses.
The problem is that the distinction is not always documented clearly, the commitments are voluntary rather than legally binding in most markets, and the precedent set by connected car data collection — where the Mozilla Foundation found all 25 major brands failing privacy standards — does not inspire confidence that those commitments will remain stable as the commercial value of emotional state data becomes clearer.
Because the commercial value is becoming very clear.
Imagine what an insurer could do with a log of your emotional states during driving. Or what a manufacturer could do with data showing that drivers of a specific model consistently show elevated stress responses in certain conditions. Or what an advertiser could do with a system that knows you're in an elevated, receptive emotional state and can push notifications to your cabin display accordingly. The technology exists. The incentive structure to use it commercially is enormous. The regulatory framework limiting that use is, in most markets, still developing.
<cite index="16-3">Privacy advocates and Consumer Reports recommend closed-loop systems that do not record or transmit video as best practice. Beyond manufacturer statements, most model-level privacy policies and technical details — what's stored, retention length, whether images are processed in-car only — are not fully documented.</cite>
Not fully documented. For a system that watches your face continuously for every minute you spend driving.
The Next Step: Voice
The camera on your face is the current frontier. The microphone pointed at your voice is the one coming next.
Voice stress analysis — the ability to detect emotional states, deception, and psychological conditions from patterns in how someone speaks rather than what they say — has been an active research area for decades. The technology's reliability in controlled conditions has improved substantially, and its integration into vehicle AI systems is a logical extension of what driver monitoring is already doing.
Your car's voice assistant already hears you. It processes what you say to execute commands. The additional step of analysing how you're saying things — the tremor in a stressed voice, the flatness in a fatigued one, the specific micro-patterns that correlate with anger or anxiety or dishonesty — is not technically distant. It is a software and ethical decision rather than an engineering one.
<cite index="10-2">Research has established that advanced driver assistance systems must account for drivers' vulnerability to accidents and recognise that systems should be designed to accommodate human error. The integration of voice and facial analysis into a unified driver state model is the logical next step for systems that already monitor physiological and behavioural indicators.</cite>
The question of whether a car should be permitted to assess whether its driver is being honest — during, for example, a voice command or a phone call conducted through the vehicle's system — is not science fiction. It is a policy question that the regulatory bodies and manufacturers will be answering in the next three to five years, whether the public is aware the conversation is happening or not.
The Safety Case Is Real. And It Is Real.
It would be dishonest to write about this technology as if safety were merely the excuse for something more sinister. The safety case is genuine, and the data behind it is solid enough that dismissing it would be irresponsible.
Driver fatigue and inattention account for approximately 25 percent of all road incidents in multiple large-scale studies — more than speeding, more than traffic violations. <cite index="12-1">More than 1.19 million people die in road accidents globally each year. Fatigue and inattention alone accounted for 25 percent of incidents in 2024 data — more than speeding at 15 percent and traffic infractions at 19 percent.</cite>
A system that detects fatigue before a driver loses consciousness. A system that notices the specific eye-movement pattern that precedes microsleep and issues a warning two seconds before the driver has lost awareness. A system that recognises the particular facial tension pattern associated with aggressive emotional states and suggests pulling over before the driver does something they'll regret. These are not trivial benefits. They save lives, and the research supporting them is among the more rigorous in applied AI.
The technology is good. The safety case is legitimate. The question is not whether the camera should exist.
The question is what it does with what it sees — and who else gets to know.
What to Ask Before You Buy
<cite index="18-1">Driver monitoring is an emerging technology currently fitted mainly to models launched since August 2024, but in a few years you won't be able to buy a new car without one. It's crucial that car makers work harder to ensure their systems work perfectly, only intervening when there is a genuine risk of an accident.</cite>
Before your next car purchase, the questions worth asking are specific: Does this system process data locally or transmit it? What is stored, and for how long? Does the manufacturer's privacy policy permit sharing driver monitoring data with insurers or third parties? Is there an opt-out mechanism, and what does opting out actually disable?
<cite index="23-1">Privacy advocates recommend confirming, in writing, each manufacturer's data collection, retention, and sharing policies before purchase — the public reporting recommends this strategy for buyers who are privacy-conscious or worried about data resale to insurers.</cite>
The camera is there. It's already watching. In most new cars sold today, it is reading your face, assessing your state, and making decisions based on what it concludes.
Whether what it knows stays between you and your car — or gets shared with someone who finds that information commercially useful — is a question worth asking before you sign.
(Image suggestions: Infrared car cabin camera close-up / Subaru DriverFocus interior camera location / Driver face with AI emotion overlay visualization / Car interior night vision camera detection / Volvo EX90 interior monitoring system)
Leave a Comment
Your comment will appear after moderation.