Inside Waymo's Driverless Safety Net: How the Company Prepares for the Moment a Computer Fails

The empty driver's seat still catches some passengers off guard. As the doors close and the vehicle begins to move without anyone at the wheel, a familiar anxiety surfaces: What if something goes wrong?

Waymo has spent years addressing that question, and now the company is pulling back the curtain on the technology designed to keep riders safe when the unexpected occurs.

The Alphabet-owned autonomous vehicle company has released its most detailed explanation yet of the computing architecture powering its growing fleet of driverless taxis. The information answers a fundamental question that has lingered over the robotaxi industry: With no human able to grab the wheel, what happens when the electronic brain running the show encounters a problem?

Built-In Redundancy

Waymo's approach centers on redundancy—multiple backup systems layered throughout the vehicle so that a single failure does not immediately create a crisis.

The vehicle carries two independent computing systems that operate simultaneously during normal driving. If one develops a fault, the other can assume control, essentially providing an electronic co-pilot already on board rather than forcing the car to wait for remote intervention.

Beyond the primary computers, Waymo has installed a secondary system specifically designed to bring the vehicle to a safe stop if it detects a failure in the main architecture. Critical functions like braking and steering each have dedicated backup hardware with separate controllers and power supplies. Should the primary braking system fail, a secondary system can still halt the vehicle. The same principle applies to steering.

Electric power follows the same philosophy. Rather than relying on a single source, critical driving systems draw from independent power supplies. A problem with one electrical component will not cascade into knocking out everything the vehicle needs to operate.

"The priority in a serious failure is to keep the vehicle under control and reach a safe state," Waymo explained in its technical documentation. That might mean stopping the trip short, but for a passenger in the back seat, a controlled stop beats the alternative.

Sensing the World

A Waymo vehicle absorbs an enormous amount of information while navigating city streets. Its sixth-generation sensor suite includes 13 cameras providing visual coverage in all directions. Lidar sensors use laser pulses to build a detailed three-dimensional picture of the vehicle's surroundings, while radar measures the distance and speed of objects around it.

All of that raw data flows into the onboard computer, which must make sense of it quickly enough to steer, brake, or react to sudden changes. Waymo emphasizes that this processing happens entirely within the vehicle itself—no connection to a distant data center, no delay waiting for cloud computing resources.

The company says it has increased raw computing power approximately 20-fold over the past eight years. That computational muscle helps shrink the time between detecting an obstacle and responding to it, a critical factor when milliseconds matter.

Waymo has also developed a custom 5-nanometer chip, technically known as an application-specific integrated circuit, or ASIC. Unlike general-purpose processors, this chip was designed for one specific job: handling the flood of information flowing in from the vehicle's sensors. The company claims it delivers more than 1,000 TOPS—trillions of operations per second—allowing the system to simultaneously process footage from all 13 high-resolution cameras while improving low-light detection.

The computing hardware connects to the vehicle's liquid-cooling system, helping maintain performance despite the harsh environment inside a working car. The computer endures constant vibration, temperature swings, and the physical demands of daily operation.

The Human Question

Waymo does employ remote workers, but the company is careful to distinguish their role from what many passengers might imagine. Remote Assistance agents provide information when the vehicle encounters unusual situations, such as asking for additional context about an ambiguous scenario. The vehicle itself decides whether to act on that information.

Remote agents do not continuously monitor every vehicle, and they do not take control of the wheel. If an onboard computer develops a hardware fault, a remote employee does not suddenly become the driver. The backup, by design, lives inside the vehicle.

This distinction matters as the industry grapples with public expectations. A driverless taxi has a fundamentally different job than the driver-assistance features available in many personal vehicles today. A typical car might warn you about a vehicle in your blind spot or help keep you centered in your lane. You remain responsible for driving. A fully autonomous Waymo must handle the complete task with no one aboard capable of taking over.

Software, Not Just Hardware

Redundancy addresses one category of risk, but it cannot solve every problem a driverless vehicle might face. A computer can function exactly as designed while its software struggles to interpret an unusual situation on the road.

Waymo learned this lesson recently. The company issued a software recall affecting 3,871 vehicles after robotaxis incorrectly entered closed freeway construction zones in Phoenix and the San Francisco Bay Area. The issue stemmed not from hardware failure but from how the automated driving system interpreted changing road conditions—a reminder that technical reliability and situational awareness are separate challenges.

The company continues refining its perception and decision-making systems as the fleet expands.

Measuring Success

Waymo's latest published safety analysis covered more than 220 million miles driven entirely autonomously through March 2026. The company claims its vehicles experienced fewer serious or fatal injury crashes than human drivers operating in comparable areas during that period. Those figures come from Waymo's own research and have not been independently verified.

The company has previously described testing scenarios involving hardware faults, power problems, and software failures. Vehicles continually check their own systems while driving, a process that becomes essential when the traditional human backup has been removed from the equation.

For the Anxious Rider

For those considering their first Waymo ride—or those who have already climbed into the back seat wondering what happens next—the company's disclosures offer a glimpse into the engineering philosophy behind the service.

If the onboard computer develops a fault, a secondary system stands ready to take over. Backup braking and redundant steering provide additional layers of protection. The empty driver's seat does not mean the absence of backup; it means the backup has been moved into the hardware itself.

Passengers who need assistance can reach Rider Support from inside the vehicle or through the Waymo app. The company says support staff can assist during a trip and in emergencies.

As autonomous vehicles become a larger presence in American cities, questions about safety architecture will only grow more important. Waymo's decision to pull back the curtain on its systems gives the public a clearer picture of how one company approaches the problem of driving without a driver—though the ultimate test remains the unpredictable, chaotic reality of actual roads.