General Motors is preparing for a moment when taking hands off the steering wheel will no longer be the most remarkable part of automated driving. The company plans to introduce “eyes-off” capability as early as 2028, beginning with the Cadillac Escalade IQ, allowing drivers to stop continuously watching the road when the system is operating within approved conditions.
The engineering challenge is considerable, but GM increasingly describes another obstacle as equally important: convincing ordinary vehicle owners that the technology deserves their confidence. After years of robotaxi controversies, confusing automation terminology and highly publicized failures across the industry, technical capability alone may not be enough. GM has accumulated enormous amounts of assisted-driving experience through Super Cruise, yet eyes-off operation crosses a psychological line. The vehicle will no longer merely help an attentive driver. For periods of the trip, the driver will be expected to trust it to handle the driving task.
The 2028 Target Is Ambitious—but It Is Not an Unconditional Promise
GM first announced its plan to bring eyes-off driving to customers in 2028, with the all-electric Cadillac Escalade IQ serving as the launch vehicle. The company has since repeated the target, and CEO Mary Barra said during GM’s second-quarter 2026 earnings call that the program remained on track and was largely meeting its milestones. The initial deployment is expected to focus on highways before expanding into more complicated driving environments.
There is an important qualification behind that date. GM describes the introduction as planned for “as early as 2028” and says deployment remains subject to development, testing, validation and other factors. That distinction matters because automated-driving timelines have repeatedly proved difficult across the industry. GM is therefore trying to establish a visible destination without pretending that software development, regulatory approval and safety validation follow a perfectly predictable calendar. For customers, 2028 is best understood as GM’s current launch target rather than a guarantee that every technical and regulatory hurdle has already been cleared.
“Eyes-Off” Represents a Much Bigger Leap Than Hands-Free Driving
Today’s Super Cruise can control steering, acceleration and braking on compatible roads while allowing the driver to remove their hands from the wheel. The important limitation is contained in the phrase “eyes on.” Drivers remain responsible for monitoring traffic and must be ready to intervene. An interior attention camera tracks head and eye position, while escalating visual and audible warnings are designed to bring an inattentive driver back into the driving task.
Eyes-off changes that relationship. GM says its planned system would allow the driver to look away from the roadway when the feature is properly engaged within its approved operating conditions. The driver would still have to remain available to take over if requested, meaning this is not the same thing as a vehicle that can drive anywhere without human involvement. That distinction closely resembles the conditional-automation concept recognized by federal safety regulators: the automated system performs the driving task while activated, but the human remains the fallback when the system requests a transition.
A Billion Super Cruise Miles Give GM a Valuable Starting Point
GM is not entering the next phase of automation with only laboratory prototypes. In April 2026, the company announced that customers had driven more than one billion miles with Super Cruise. Nearly 750,000 Super Cruise-enabled vehicles across 23 North American models had contributed to that total, giving GM a substantial installed base from which to study how automated assistance behaves outside controlled development environments.
Customer usage also provides clues about acceptance. GM reported that owners had used Super Cruise for 7.1 million hours across 28.7 million trips during the preceding 12 months. More than half of Super Cruise drivers were using it weekly, while nearly 85% activated it at least once a month. Those figures do not prove customers will automatically embrace eyes-off driving, but they provide GM with something an entirely new entrant would lack: hundreds of thousands of people already accustomed to handing portions of the driving task to software. The next challenge is persuading them to surrender continuous visual supervision as well.
GM Is Trying to Test the Situations Drivers Rarely Think About
Autonomous systems are relatively easy to demonstrate when roads are dry, markings are clear and surrounding drivers behave predictably. The difficult problems live in what engineers call the “long tail”: unusual construction layouts, abrupt weather changes, confusing human behavior and uncommon combinations of circumstances that may occur only rarely but still have to be handled safely. GM says those edge cases are central to its validation strategy.
The company is combining three major sources of information: real-world telemetry from production vehicles, high-precision data from development fleets and synthetic scenarios created in simulation. GM has said its simulation systems can reproduce roughly 100 years of driving every day, allowing engineers to replay unusual situations and alter variables repeatedly without waiting for the same event to happen again on public roads. Supervised public-road testing is also underway, including development work in California and Michigan. The scale is impressive, but the more important question is whether those billions of virtual and real miles adequately represent the rare situations that determine public confidence after launch.
Cameras Alone Are Not GM’s Answer to the Safety Problem
GM’s planned eyes-off system is being developed around multiple sensor types rather than depending on a single form of perception. The company says redundancy will include LiDAR, radar and cameras integrated into the vehicle. LiDAR can construct detailed three-dimensional information about the environment, radar performs particularly useful distance and velocity measurements, and cameras provide visual information such as lane markings, signs and object classification.
Behind those sensors, GM is also preparing a major computing overhaul. Its second-generation software-defined vehicle architecture is scheduled to arrive in 2028 alongside the eyes-off system, beginning with the Escalade IQ. GM says the centralized platform will connect major systems including propulsion, infotainment and safety through a high-speed computing core, providing 10 times greater over-the-air update capacity, 1,000 times more bandwidth and up to 35 times more AI performance than its previous architecture. Those numbers sound like technology specifications, but their real significance is redundancy and response time: a vehicle entrusted with the complete driving task needs enough sensing and computing capacity to recognize problems and respond reliably.
Cruise’s Difficult History Is Now Part of GM’s Autonomous-Driving Strategy
GM’s road to personal autonomy runs directly through Cruise, the robotaxi company it backed for years. In December 2024, GM announced that it would stop funding Cruise’s standalone robotaxi development and instead combine the technology and engineering expertise with its own driver-assistance work. GM completed its acquisition of Cruise’s remaining ownership in February 2025, turning the operation into a wholly owned business focused more directly on personal vehicles.
That strategic retreat did not mean the technology disappeared. GM says Cruise contributed more than five million miles of fully driverless experience, along with perception technology, AI development and simulation systems that are now feeding its next-generation automated-driving work. The history also provides a cautionary lesson. Building impressive autonomous prototypes is different from operating technology safely, predictably and transparently at scale. GM’s current strategy effectively takes the technical knowledge accumulated during the robotaxi era and places it inside vehicles sold directly to customers. That shifts the trust relationship from passengers trying a service to owners relying on the system repeatedly for years.
Consumer Confidence Remains Far Behind the Technology
Automakers can demonstrate increasingly sophisticated autonomous systems, but public attitudes have moved much more slowly. J.D. Power’s 2026 U.S. Mobility Confidence Index, conducted with the MIT Advanced Vehicle Technology Consortium, found that fewer than one in four consumers felt comfortable riding in a fully self-driving vehicle. The overall confidence index remained at 39 out of 100, essentially unchanged from 2024.
Safety remained the largest obstacle. Sixty percent of respondents identified personal safety as a leading concern, 58% were worried about how autonomous vehicles would handle emergencies, and 51% questioned performance in difficult conditions such as heavy traffic or bad weather. Earlier AAA research produced a similar message: only 13% of U.S. drivers surveyed in 2025 said they would trust riding in a self-driving vehicle, while roughly six in ten said they were afraid. GM’s eyes-off system will not be identical to a fully driverless robotaxi, but those attitudes show the environment into which it will arrive. Trust cannot be assumed simply because a system performs well technically.
The Moment the Car Gives Control Back May Be the Hardest Part
Allowing someone to look away from the road creates a human-factors problem that does not exist in the same way with today’s Super Cruise. A driver watching traffic can react almost immediately when automation disengages. A driver reading, working or focusing elsewhere must first recognize the request, understand the roadway situation and mentally rebuild awareness before deciding what action to take.
Research into conditional automation has repeatedly found that non-driving activities can affect takeover performance. A 2024 meta-analysis examining dozens of studies found that non-driving tasks could produce longer reactions and poorer vehicle-control performance during transitions back to manual driving. The problem becomes especially important when cognitive attention is deeply occupied elsewhere. That does not mean eyes-off automation is inherently unsafe; it means successful systems must be designed around realistic human behavior rather than assuming perfect responses. Alerts, transition timing, fallback procedures and restrictions on what occupants can safely do while automation is active could therefore matter almost as much as how accurately the vehicle stays in its lane.
Regulators Will Expect Evidence, Not Just Impressive Demonstrations
Eyes-off capability also moves GM into a more demanding regulatory category. NHTSA distinguishes today’s Level 2 driver-assistance technology from automated driving systems covering SAE Levels 3 through 5. Level 2 requires the human to continuously monitor driving, while higher automation can perform the complete dynamic driving task within defined conditions. That means a future GM system that no longer depends on continuous driver vigilance will receive different scrutiny from ordinary highway assistance.
Federal oversight is already evolving. NHTSA requires manufacturers and operators to report certain crashes involving both automated driving systems and Level 2 assistance technologies, allowing regulators to investigate potential defects and emerging safety patterns. The agency has also continued developing its broader automated-vehicle framework. For GM, trust will therefore have two audiences: customers and regulators. A vehicle may perform flawlessly during thousands of routine journeys, but a serious failure, unclear system boundary or badly handled takeover request could become a regulatory issue quickly. Transparency about where the technology works—and where it does not—will be essential.
Trust Will Probably Be Won During Ordinary Drives, Not Spectacular Demos
GM product manager John Kaychi has summarized the problem plainly: even excellent technology will struggle if customers do not trust it. His team is emphasizing consistent performance and a domain-by-domain rollout rather than attempting to make the first version work everywhere. GM currently intends to begin with highways, where traffic flows are comparatively structured, before eventually moving toward broader driveway-to-driveway capability.
That restrained approach could prove important. Drivers are unlikely to develop confidence because of a technical presentation explaining sensor fusion or artificial intelligence. Trust is more likely to emerge after hundreds of uneventful lane changes, smooth responses to merging traffic, understandable warnings and predictable decisions in rain, construction or congestion. Conversely, a system that frequently surprises its owner may lose confidence even if its overall statistics appear impressive. GM already has considerable experience persuading customers to take their hands off the wheel. By 2028, it hopes to persuade them to look away as well. The harder achievement may not be making the vehicle capable of doing it, but making that decision feel routine rather than reckless.
