On August 20, General Motors’ own newsroom published a nostalgic technical retrospective crediting the doomed 1990s EV1 with inventing brake-by-wire, a system GM once called the world’s most efficient and intelligent stopping system. One day later, federal regulators expanded a defect investigation into 1.16 million GM-built vehicles running the modern descendant of that same technology. Nobody at GM planned that collision of timelines. But it is the most useful coincidence in automotive journalism this year, because it points at what actually matters: this isn’t really a story about a fractured spindle in a Cadillac. It’s a story about what GM gave up when it decided a computer, not a rubber diaphragm, would decide how a car stops.
Every gas car built before the EV1, and plenty built since, stops the same simple way. Push the pedal, hydraulic fluid moves, a vacuum booster amplifies the driver’s leg force, and pads clamp onto rotors. A vacuum booster is a rubber diaphragm and a check valve. It has no software, no sensors, and almost nothing that fails halfway. It works, or it doesn’t, and a competent independent shop can diagnose one with a hand pump.
GM’s engineers ran into that simplicity as a problem when they built the EV1. An electric motor can act as a generator, feeding energy back into the battery as the car slows down, which is worth real range on a car that only had about 90 miles to give. But regenerative braking and a purely mechanical vacuum booster can’t share one brake pedal. Something has to decide, instantly, how much stopping power comes from the motor and how much comes from the pads, then blend the two without the driver feeling the handoff. GM’s answer was to put a computer between the pedal and the brakes, a system it called brake-by-wire, run by what it named the Brake Torque Control Module.
Here is the detail worth sitting with. GM’s own account of the EV1 makes a point of noting that, for safety, the brake pedal is also connected to a fully mechanical backup system behind all that computing. In 1996, even GM didn’t trust software alone to stop a car. That detail is in the company’s proud version of its own history. It is not in the explanation GM gave NHTSA about how a fractured spindle takes down anti-lock brakes, stability control and traction control in the modern eBoost system. Maybe an equivalent mechanical fallback still exists in today’s design. But if it does, GM hasn’t described one in the filing where it’s supposed to explain exactly how its brakes fail.
The EV1’s control logic used what GM calls a friction-first strategy: press the pedal, and the car engages the mechanical brakes immediately, then blends in regenerative braking based on how much charge the battery can accept. Brandon Vivian, now executive chief engineer for GM Defense and Cadillac V-Series, worked on that system starting in 1995 and later moved to the Precept concept, which flipped the logic to regen-first, decelerating electrically first and adding friction braking only as needed. Vivian says that is the basic strategy GM still uses today. It is also, not coincidentally, exactly the kind of decision-tree software logic that is nearly impossible to test against every real-world combination of speed, battery state and component wear, and easy for a failure to slip through in a way no lab predicted.
What makes the current investigation bigger than a Cadillac problem is that eBoost, GM’s modern brake-by-wire hardware, stopped being an EV-only part years ago. It’s standard equipment on the gas-burning Chevrolet Colorado and GMC Canyon, trucks with no regenerative braking to blend in at all. GM uses it there because a computer-mediated brake pedal can be tuned for feel, Corvette and Cadillac V-Series models change how firm the pedal feels depending on drive mode, and because features like Super Cruise and automatic emergency braking need a brake system a computer can actuate on its own, not merely amplify. eBoost, in other words, isn’t brake technology GM built for electric cars. It’s brake technology GM built so software could get involved in stopping every kind of car, and EVs just happened to go first.
That shift matters to anyone who owns one of these vehicles, recall or not. A worn vacuum booster is a bounded, well-understood failure that most brake shops can diagnose and replace without a manufacturer’s scan tool. eBoost is a motor, a control module and a set of diagnostic trouble codes that mostly only make sense to a dealer’s software. When it fails, GM’s own filing describes a cascading shutdown: warning chimes, a Service Brake System message, and a speed limiter capping the car at 43 mph. That reads less like a blown fuse and more like a computer politely giving up. It’s the tradeoff nobody puts in a heritage post: the same complexity that lets a Corvette’s brake pedal feel different in Track mode is the complexity that turns a simple fix into a dealer-only, sensor-replacement job.
None of this makes GM uniquely careless. Brake-by-wire is where the entire industry is headed, for the same reasons GM chased it in 1996: it’s the only way to blend regenerative braking, and it’s increasingly required to run the driver-assistance features regulators and customers now expect. Our own reporting on GM’s Super Cruise expansion into Canada and its Ultium battery manufacturing strategy both trace back to the same underlying bet, that software-mediated systems, built once and deployed everywhere, are how a car company survives the next decade. The eBoost investigation, which we covered in detail when NHTSA expanded it to 1.16 million vehicles, is what that bet looks like when the software’s model of its own failure mode doesn’t quite match what’s happening on the road.
A vacuum booster either works or it doesn’t. Brake-by-wire negotiates, between the motor and the pads, between the driver’s intent and the battery’s state of charge, between what engineers modeled and what thirty years of real driving eventually finds. That negotiation is why GM could brag, accurately, that the EV1 pioneered smarter brakes. It’s also why an engineering analysis with GM’s name on it exists at all three decades later. The EV1 sold about 1,100 copies and never once needed a fleet-wide safety filing. Its great-grandchild is on the road 1.16 million times over, and regulators are still trying to find out whether the computer between the pedal and the pads always makes the decision GM says it does.