GM’s Eyes-Off Push Puts It Directly Against Tesla and Mercedes on the Autonomous Trust Problem
General Motors has drawn a clear line in the autonomy race. This week, the company confirmed that 200 test vehicles are now running on public highways in California and Michigan as part of supervised testing for its eyes-off driving system — a technology it plans to launch in 2028, starting with the all-electric Cadillac Escalade IQ. The milestone puts GM on a direct collision course with Tesla’s Full Self-Driving and Mercedes’ Drive Pilot, two systems that have generated headlines but left a trust gap that GM believes it can close.
John Kaychi, GM’s Senior Group Product Manager for Automated Driving, has been explicit about the company’s bet. Technical capability is table stakes. What actually determines whether a mass-market autonomy product succeeds, he argues, is whether everyday drivers feel confident enough to use it. “You can have the best technology in the world,” Kaychi said. “But if customers don’t trust it, it won’t stick.”
What Eyes-Off Actually Means, and Why the Bar Is Higher Than You Think
The distinction between GM’s current Super Cruise system and the 2028 eyes-off capability is significant. Super Cruise is a Level 2 system: hands can leave the wheel, but the driver must remain visually engaged with the road. Eyes-off is a different category entirely — under defined conditions, drivers can legitimately divert their attention to something else. That pushes the system toward Level 3 autonomy, where the car, not the driver, is responsible for monitoring the environment.
Mercedes’ Drive Pilot already holds Level 3 certification in the U.S., but its operational envelope is narrow — limited to specific mapped highways at low speeds. Tesla’s FSD, despite its name, remains a Level 2-plus system that legally requires the driver to stay attentive and ready to intervene at any moment. A crash in Issaquah earlier this summer, where a Tesla operating in FSD mode swerved into oncoming traffic and had its rear section sheared off, served as a sharp reminder of the gap between marketing language and real-world accountability. No injuries were reported, but police reiterated that FSD does not transfer responsibility from the driver.
GM’s Three-Pillar Validation Strategy
Rather than relying on any single data source, GM is validating its eyes-off system against three distinct streams. The first is telemetry from its existing production fleet — millions of Super Cruise miles already logged by retail customers across varied real-world conditions. The second is data from a dedicated next-generation development fleet, equipped with high-precision sensors designed to capture the kind of granular detail that standard production hardware misses. The third is synthetic testing: high-fidelity simulated environments built specifically to stress-test rare edge cases that real-world miles alone can’t reliably generate.
GM says the system had already been trained on millions of real-world miles and stress-tested millions of times in simulation before this week’s public road testing began. The 200 test vehicles currently running on California and Michigan highways each carry a safety driver capable of taking control, but the program represents the first time GM’s next-generation system has operated on public limited-access roads. From there, the rollout plan moves domain by domain — highways first, then driveway-to-driveway capability — validating each expansion before opening the next.
Kaychi’s Robotaxi Background Shapes GM’s Retail Approach
Kaychi’s read on the trust problem isn’t theoretical. Before joining GM, he worked at Zoox, the Amazon-owned robotaxi company, where he helped deploy the company’s first employee ridesharing service on San Francisco’s most demanding streets — steep grades near Coit Tower, blind crests, dense urban corridors. Watching people experience autonomous vehicles for the first time in genuinely complex conditions gave him a specific frame for what the retail challenge looks like at scale.
“Getting to see the technology through the eyes of so many people who hadn’t experienced it before reminded me how groundbreaking this technology is,” Kaychi said. That experience now informs how GM thinks about the gap between a controlled demonstration and a product that has to build confidence through consistent daily performance — not just impress someone once.
For the 2028 launch, GM is combining two internal assets: Super Cruise’s proven track record of scaling driver-assistance technology across a mass-market customer base, and Cruise’s expertise in AI development and edge-case testing. The centralized computing architecture underpinning the system is also designed to scale — GM says eyes-off capability will eventually extend from premium Cadillacs down to mainstream Chevrolet models without rebuilding the system from scratch for each vehicle.
The autonomy race has never been short on technical ambition. What’s been harder to demonstrate is the kind of consistent, predictable real-world performance that makes a driver genuinely comfortable looking away from the road. GM’s 2028 timeline is aggressive, but the deliberate validation approach — and the choice to start with a flagship like the Escalade IQ — suggests the company is more focused on getting the trust equation right than on being first.
Related posts
Latest Post
No Comments