ADAS and partial automation: how to use Level 2 systems safely
What Level 2 systems can do, where they fail and when to take back control
Picture a straight interstate at night. Your car is steering itself behind a pickup when the pickup swerves into the next lane, revealing a stalled car in yours. Your system may not react in time, so that moment is yours to handle.
Systems that steer and control speed at once, some hands free, are what NHTSA calls Level 2. They support the driver. NHTSA states that no vehicle currently for sale is fully automated [2].
Here is a human factors engineer’s guide to its limits, its driver monitoring and the habits that keep you in control. Beginners should start with our guide to driver assistance technology for new drivers.
Driver support vs. automated driving: the levels in plain words
Engineers rank automation on a 0 to 5 scale from SAE International. NHTSA explains it this way [1, 2]:
- Level 0: Brief help, such as warnings and automatic emergency braking (AEB).
- Level 1: Ongoing help with steering or speed, but not both, such as adaptive cruise control (ACC).
- Level 2: Ongoing help with steering and speed together, while you stay fully engaged and attentive.
- Level 3: The system drives in limited conditions. You must be ready to take over when it asks. NHTSA says this is not widely available to buy.
- Levels 4 and 5: The system does all the driving, in limited service areas (Level 4) or everywhere (Level 5). Neither is for sale to consumers.
The big line falls between Levels 2 and 3. Through Level 2, NHTSA’s message is that you do the driving and the watching. You must steer, brake and accelerate [1].
Level 3 in the U.S. is narrow. California’s Department of Motor Vehicles (DMV), for example, permits a Mercedes-Benz system only on certain freeways, in daytime, at up to 40 mph. Heavy fog, smoke, dust and snowstorms are excluded [8]. Robotaxis are a separate category: Level 4 services operating in limited areas, not cars you can buy [2].
How Level 2 works: hands on and hands free designs
A Level 2 system pairs ACC, which keeps a set gap behind the vehicle ahead, with lane centering, which steers to keep you centered [1]. Designs fall into two groups:
- Hands on systems expect a hand on the wheel. Some check it by sensing light steering force, called steering torque.
- Hands free systems allow hands off the wheel on approved roads while a camera watches your eyes and head. One system the National Transportation Safety Board (NTSB) studied allowed this only on certain mapped, limited access highways [3].
Each system has an operational design domain (ODD): the road types, speeds, lighting and weather it was built for. Since 2017, NTSB has urged NHTSA to make sure Level 2 systems can be used only in the conditions they were designed for [9].
At a glance: each feature, its limits and your job
Features vary by model, trim and software version, so check your owner’s manual.
Comparison compiled by Safety Behind the Wheel Foundation
| Feature | What it does | Known limits | Your job |
|---|---|---|---|
| Automatic emergency braking | Brakes, or adds braking, when a front crash is imminent [1] | Failed to activate for stopped cars at night at highway speed in 2 crashes NTSB studied; missed a highway crash barrier in another [3, 9] | Brake for what you see; never wait for the car |
| Lane keeping assistance | Gently steers to keep you from drifting out of your lane [1] | Can struggle on poorly marked or snow covered roads [5] | Steer yourself; treat repeated alerts as a sign of fatigue or distraction |
| Lane centering | Steers continually to hold the middle of the lane [1] | Steered into the wedge at a highway exit in one fatal crash [9]; sensors may falter in low light or bad weather [5] | Watch where the lane goes, not just where the car goes |
| Adaptive cruise control | Adjusts speed to keep a preset gap to the vehicle ahead [1] | One system was designed not to slow for stopped vehicles above 65 mph [3] | Choose a safe speed and gap; cover the brake when traffic bunches up |
| Level 2, hands on | Combines ACC and lane centering; expects hands on the wheel [1] | Steering torque checks can’t tell where your eyes are [9] | Eyes on the road, hands on the wheel, ready to override |
| Level 2, hands free | Same, with hands off on approved, mapped highways [3] | One camera system missed repeated short glances away and a phone held up in front of the driver [3] | Treat hands free as hands ready and eyes on |
| Driver monitoring | Checks hands, eyes or both, and escalates alerts when you’re inattentive [6] | No U.S. performance standard, NTSB noted in 2026 [3] | Respond to the first alert; don’t learn to game it |
Where Level 2 struggles
In AAA’s 2022 track tests, three Level 2 vehicles avoided a slower car and a cyclist ahead in every run [4]. The trouble lies in situations that are rare on any one trip but likely over years of driving.
Stopped vehicles at speed. NTSB investigated two fatal 2024 nighttime crashes in which SUVs using a hands free system struck stopped vehicles on interstates [3]. The adaptive cruise was designed not to slow for stationary vehicles above 65 mph, to avoid false alarms. AEB did not activate [3]. A system tuned to ignore false alarms can also ignore a real hazard.
Oncoming, crossing and merging traffic. In the AAA tests, an oncoming car crossing into the lane caused a crash in all 15 runs. A crossing cyclist was hit in 5 of 15 [4]. In a AAA Foundation for Traffic Safety study, drivers were told the system could not detect vehicles moving into their lane or construction workers [7].
Work zones. In one NTSB case, hands free mode stayed available in a work zone whose lowered 45 mph limit was missing from the car’s map data. That driver was also impaired and going about 27 mph over the limit [3].
Exits and roadside hardware. In a 2018 crash, a Level 2 system steered into the paved wedge where an exit splits off. Its collision systems missed the crash barrier there [9].
Lines, light and weather. IIHS notes that sensors may not work well in low light or bad weather. Lane systems can struggle on poorly marked or snow covered roads [5]. A low sun can dazzle cameras. Dirt or ice can block sensors. Sharp curves may need more steering than a system will give.
Driver monitoring: how the car checks on you
Steering torque sensing detects pressure on the wheel, which NTSB calls an ineffective measure of engagement [9]. In the 2018 crash, the car sensed no steering torque during 34.4% of the time the system was on [9]. A hand on the wheel also says nothing about where your eyes are.
Camera based monitoring, which tracks your gaze and head position, is a step up. But NTSB found gaps in one 2024 system. Its first alert came after 5 seconds of looking away. Brief glances at the road reset that timer. It also couldn’t tell the road from a phone held in the driver’s line of sight [3]. By our calculation, a car at 70 mph covers about 513 feet in 5 seconds, more than 1.5 football fields.
How good systems escalate. IIHS criteria call for two types of alerts within 10 seconds of inattention, and a third type or a slowdown before 20 seconds [6]. A slowdown should begin within 35 seconds of disengagement, followed by a lockout for the rest of the drive [6].
In its first round, in March 2024, IIHS rated 1 of 14 systems acceptable, 2 marginal and 11 poor. None met all its driver monitoring criteria [6]. Its ratings list still showed those 14 systems in October 2026 [10]. Ratings apply to specific software versions. IIHS noted that two systems it rated ran software from before a December 2023 recall [6].
People also adapt. In an IIHS and MIT AgeLab study, 14 first time users of one system learned its alert pattern within a month. They answered alerts faster but spent more time on nondriving activities [11].
NTSB asked NHTSA in 2020 for Level 2 driver monitoring standards [9], and in 2026 to require effective monitoring on all new vehicles with partial automation [3].
Does Level 2 make driving safer? What the data shows
IIHS found no crash reduction advantage for vehicles with partial automation over same brand vehicles with only crash avoidance features. It also reports that regular users often have a false sense of security [5].
Under NHTSA’s Standing General Order (SGO), in place since 2021, named companies must report serious crashes when Level 2 was in use within 30 seconds beforehand. That covers crashes with a death, hospital transport, air bag deployment or struck pedestrian or cyclist [12].
The data help NHTSA spot defects, but counts aren’t adjusted for each company’s vehicles or miles [12]. Companies with better connected car data may report more crashes simply because they learn of more. Reports can be unverified or duplicated [12]. So you can’t rank brands by SGO counts. NTSB also found the automaker learned of each 2024 crash weeks later, once from an online forum and once from NTSB investigators [3].
A recall can change how your system works, often through a software update. Our guide to vehicle recalls and safety defects explains how to check your vehicle identification number (VIN).
Why attentive drivers drift out of the loop
Automation complacency. When a system works most of the time, people stop checking it. In an IIHS survey of 604 regular users of three Level 2 systems, all groups did more nondriving activities with the system on. Depending on the system, 12% to 53% felt comfortable treating it as fully self driving [13].
Vigilance decrement. Watching for rare problems wears you down. In a 40 minute simulator study, all drivers spotted fewer hazards over time. Only those whose car held speed and lane also got worse at telling real hazards from harmless events [14].
Mode confusion. Drivers can lose track of what the car is doing, such as when lane centering drops out but ACC stays on. In a NHTSA test track study, a Level 2 prototype drifted out of its lane while drivers did a visual task. With no alert, 23 of 49 drivers had to be told to take control, compared with 1 of 54 who got an alert [15].
Slower takeovers. In the AAA Foundation study, 23% of drivers taught to see a system as highly capable took over 5 seconds to grab the wheel when it quit. Only 6% of drivers taught its limits were that slow [7]. In NHTSA’s Level 3 test, drivers took 17 seconds on average to respond to a gradual, nonurgent handback [15]. Many drivers seem to wait until the car insists.
Marketing names. In that AAA Foundation study, both groups used the same system. One learned it as “AutonoDrive,” stressing convenience, and the other as “DriveAssist,” stressing limits. The AutonoDrive group overestimated its abilities and spent more time with hands off the wheel [7]. A name does not tell you what a system can do. The owner’s manual does.
Before you engage Level 2: a checklist
For choosing a gap, see our guide to safe following distance.
When to switch it off, and how to take back control
Turn Level 2 off, or leave it off, when:
- You see work zone signs, cones, flaggers or lane shifts.
- Rain, snow, fog or glare limits your view. Assume the sensors struggle too.
- You leave the highway for streets with intersections, pedestrians and cyclists.
- Lane lines are faded or confusing, or the road is unfamiliar.
- You’re near exits, lane splits, toll plazas, emergency lights or a crash scene.
- Attention alerts keep coming, which can signal fatigue.
When the car asks you to take over:
- Eyes up first. Look far ahead and check your mirrors.
- Hands on, foot ready. Grip the wheel and cover the brake.
- Drive it yourself. Steer and slow as needed. Don’t wait on the system.
- Confirm the mode. Glance at the display to see what is still on.
- Keep it off if unsure. If you don’t know why it handed back control, drive manually the rest of the trip.
- Report patterns. Tell your dealer, and file a safety complaint with NHTSA if a problem repeats.
Treat a slowdown or lockout after ignored alerts as a firm warning. New drivers should keep Level 2 off until core skills are solid. Employers should apply normal phone and rest rules to Level 2 time.
Myths vs. facts
Frequently asked questions
Is Level 2 the same as a self driving car?
No. You are still the driver [1]. NHTSA says no fully self driving vehicle is for sale today. Level 3 permits so far are narrow [2, 8].
Can I take my hands off the wheel with a hands free system?
Only where the system allows it, typically on approved, mapped highways [3]. Keep your eyes on the road and your hands close enough to steer at once.
Why does my car say I’m not paying attention when I am?
Monitoring systems judge you indirectly, from steering pressure or where your eyes and head point. They can misread you [3, 9]. Check your manual for how yours works, and make sure nothing blocks the camera.
Should I use Level 2 in rain or snow?
It’s best not to. Sensors may not work well in bad weather. Lane systems struggle when markings are covered [5]. Drive yourself and leave extra space.
Who is at fault if my car crashes while Level 2 is on?
That depends on the facts and state law, but NHTSA says the driver is responsible at Levels 0 to 2 [1]. For a specific case, ask a lawyer or your insurer.
The bottom line
Recommendation from Safety Behind the Wheel Foundation
Level 2 can ease long highway drives, but it is driver support, not a driver. Learn where your system is built to work. Keep your eyes on the road and hands ready. Switch it off the moment conditions get complicated.
Related articles
- Driver assistance technology for new drivers: what it can and can’t do
- Safe following distance: the 3 second rule and when to add more
- Vehicle recalls and safety defects: how to check, fix and report
- Work zone and move over safety: protect road crews, responders and you
- Drowsy driving: warning signs, hidden risks and what really works
Sources and further reading
- 1National Highway Traffic Safety Administration (NHTSA). Driver Assistance Technologies. No publication date shown (page cites 2024 crash deaths); accessed October 2026. nhtsa.gov/vehicle-safety/driver-assistance-technologies
- 2NHTSA. Automated Vehicle Safety. No publication date shown (page references the 2025 USDOT automated vehicle framework); accessed October 2026. nhtsa.gov/vehicle-safety/automated-vehicles-safety
- 3National Transportation Safety Board (NTSB). Fatal Crashes Between Vehicles Operating in Hands-Free Partial Automation Mode and Stationary Vehicles in San Antonio, Texas, and Philadelphia, Pennsylvania. Highway Investigation Report NTSB/HIR-26-02. Adopted March 31, 2026. ntsb.gov/…/HIR2602.pdf
- 4AAA. Consumer Skepticism Toward Active Driving Features Justified (news release on closed-course testing of three Level 2 systems). May 12, 2022. newsroom.aaa.com/2022/05/consumer-skepticism…
- 5Insurance Institute for Highway Safety (IIHS). Advanced driver assistance (research area overview). Updated September 2026. iihs.org/research-areas/advanced-driver-assistance
- 6IIHS. First partial driving automation safeguard ratings show industry has work to do (news release). March 12, 2024. iihs.org/news/detail/first-partial-driving-automation-safeguard-ratings…
- 7AAA Foundation for Traffic Safety. Singer J, Jenness JW. Impact of Information on Consumer Understanding of a Partially Automated Driving System. September 2020. aaafoundation.org/impact-of-information…
- 8California Department of Motor Vehicles. Autonomous Vehicle Permit Holders (deployment permit conditions). Updated September 18, 2026. dmv.ca.gov/…/autonomous-vehicle-testing-permit-holders
- 9NTSB. Collision Between a Sport Utility Vehicle Operating With Partial Driving Automation and a Crash Attenuator, Mountain View, California, March 23, 2018. Highway Accident Report NTSB/HAR-20/01. Adopted February 25, 2020. ntsb.gov/…/HAR2001.pdf
- 10IIHS. Partial automation safeguards (ratings list). Accessed October 2026. iihs.org/ratings/partial-automation-safeguards
- 11Mueller AS, Gershon P, Haus SH, Cicchino JB, Mehler B, Reimer B (IIHS and MIT AgeLab). Finding windows of opportunity: how drivers adapt to partial automation safeguards over time. Transportation Research Part F. May 2025. IIHS bibliography record. iihs.org/topics/bibliography/ref/2312
- 12NHTSA. Standing General Order on Crash Reporting. Third amended order effective June 16, 2025; data through August 17, 2026 shown when accessed October 2026. nhtsa.gov/laws-regulations/standing-general-order-crash-reporting
- 13Mueller AS, Cicchino JB, Calvanelli JV Jr (IIHS). Habits, attitudes, and expectations of regular users of partial driving automation systems. Journal of Safety Research. February 2024. IIHS bibliography record. iihs.org/topics/bibliography/ref/2261
- 14Greenlee ET, DeLucia PR, Newton DC. Driver Vigilance Decrement Is More Severe During Automated Driving than Manual Driving. Human Factors. Published online May 27, 2022. journals.sagepub.com/doi/10.1177/00187208221103922
- 15NHTSA. Blanco M, et al. Human Factors Evaluation of Level 2 and Level 3 Automated Driving Concepts. DOT HS 812 182. August 2015. nhtsa.gov/…/812182_humanfactorseval-l2l3-automdrivingconcepts.pdf
This article reflects the safety priorities of automotive human factors and vehicle safety engineering. It is general education, not legal or professional advice about any specific vehicle or driver. Vehicle features, software, ratings, rules and data change, so check your owner’s manual and official sources such as NHTSA. Reviewed October 2026.
Keep learning
Safe Driving Resource Center
Guides, self-checks and articles for drivers, parents of new drivers, and families recovering after a crash.
