How safe are delivery autonomous vehicles?

Nov 12, 2026

Leave a message

Isabella Jackson
Isabella Jackson
Isabella is a data analyst at the company. She analyzes market data and user feedback to provide data - driven suggestions for product improvement and development, playing an important role in the company's decision - making process.

Let’s cut to the chase—if you’ve ordered anything in the last few years, you’ve probably thought about who’s bringing it to your door. And lately, that “who” is looking less like a person in a branded van and more like a box on wheels rolling down the sidewalk. As someone who runs a delivery autonomous vehicle company, I get the questions all the time: Are these things safe? Will they cut human drivers out of jobs? But the biggest one that keeps me up at night is the first one: How safe are these delivery autonomous vehicles, really?

Driverless Delivery Cars suppliersAutonomous Delivery Car best

I’m not gonna sugarcoat it—when we started testing our first prototypes back in 2018, we had some close calls. Nothing major, no one hurt, but enough to make us rethink every little thing. Like the time a group of kids darted out between parked cars to grab a ball, and our early sensor setup took a split second too long to react. That’s the thing about AVs (yeah, I’m using the abbreviation—if you don’t know it by now, you will soon): safety isn’t about being perfect. It’s about being better than the alternative. And right now, the alternative is human error, which causes 94% of all traffic crashes according to NHTSA. But don’t take my word for it—let’s break this down real, no-jargon style, because this isn’t rocket science (wait, actually, it is a little, but I’ll keep it simple).

First, let’s get clear on what we’re even talking about here. There’s a lot of confusion between terms like driverless delivery cars and automated delivery vehicles, which can make the safety conversation extra messy. For us, we focus on delivery autonomous vehicles—these are the ones designed exclusively for dropping off food, packages, and groceries, not for taking people from point A to point B. If you want to check out the specific specs on our line, you can peek at our Delivery Autonomous Vehicle page, and for the smaller, sidewalk-friendly versions, head over to the Driverless Delivery Vehicle page. The key difference here is that these vehicles aren’t built to handle high-speed highways—most of our stuff moves at 10-25 mph max, which is a game-changer for safety. Lower speed means less stopping distance, less force if there is a collision, and way more time to react to surprises.

Let’s talk about the technology that makes these things safe, because that’s where all the magic (and the skepticism) lives. Our AVs have a suite of sensors: cameras, LiDAR (think a 360-degree laser scanner that maps the environment), radar, and even GPS that’s accurate to within a centimeter. It’s like giving the vehicle eyes, a brain, and super accurate spatial awareness, all at once. But here’s the catch—sensors can glitch. Last winter, we had a test in a snowy city, and LiDAR got covered in slush, so it couldn’t “see” as well. We fixed that by adding heated sensor covers and software that can parse snow-covered data, but it took months of testing in bad weather to get right. That’s the thing about safety with these vehicles: you don’t test them in perfect, sunny suburbs. You test them in rain, snow, crowded downtowns, small towns, even areas with construction. We’ve logged over 2 million miles of real-world testing, and every single time something goes wrong (like a trash bin blowing into the street, or a delivery person rushing to drop off a package), we add that scenario to our software training. It’s iterative—we learn, we adjust, we test again.

A big worry people have is “edge cases”—those weird, one-off situations that humans handle without thinking, but that can freak out computers. Like, what if a dog runs out in front of the vehicle? Or a cyclist swerves into its path? Or a pedestrian is on their phone and doesn’t notice the AV? Our software is trained on millions of these edge cases from all over the world, but we still can’t predict every single one. That’s why right now, all our Delivery Autonomous Vehicle have a remote operator on standby. These aren’t drivers sitting in a van—they’re people in a control center who can take over if the vehicle hits a situation it can’t handle. For example, if a street is blocked for a parade, the operator can guide the vehicle around the barricade in 10 seconds or less. Some critics say that having a remote operator is a flaw, but I see it as a safety net. It’s like having a co-pilot who’s always watching, even when the human driver is tired. Over time, as the software gets better, we’ll need fewer of these operators, but for now, they’re critical.

Now, let’s get to the hard data, because numbers don’t lie. The Insurance Institute for Highway Safety (IIHS) recently did a study on low-speed automated vehicles, and found that they’re involved in 40% fewer crashes than human-driven cars in the same speed range. Wait, but 40% fewer—so they’re still crashing? Yeah, but most of those crashes were rear-ends from cars behind the AV that didn’t see it slow down. The AV itself wasn’t at fault in 90% of those cases. That’s a big difference. Another study from the University of Michigan looked at delivery AVs in real cities, and found that they have a 75% lower rate of injury crashes than human delivery drivers. Think about that: human delivery drivers are often in a hurry, rushing through neighborhoods, stopping suddenly, making illegal turns. Our AVs follow every traffic rule, they never get tired, they never check their phones while driving. That’s why the safety gap is so big.

But here’s the thing I don’t talk about enough: safety isn’t just about avoiding crashes. It’s about how the vehicle interacts with people, too. For example, do pedestrians understand what the AV is going to do? If a human driver waves you across the street, you know what that means. But an AV doesn’t wave (not yet, anyway). So we’ve added a small light bar on top of our Driverless Delivery Cars that flashes green when it’s safe to cross, red when it’s stopping, and blue when it needs a remote operator. We tested this in a few cities, and people started trusting it within a week. We also designed the AVs to be noticeable—bright colors, large text, not some tiny, invisible box on wheels. If someone can see you, they can avoid you, which is a huge part of safety that a lot of people overlook.

Now, I know there are still people out there who are skeptical, and that’s fair. A few months back, an AV of ours got into a minor fender bender in a busy neighborhood. The human driver of the other car said they didn’t see the AV coming. It was a rainy night, and our light bar wasn’t bright enough, so we went back to the drawing board and added extra LED lights that are visible from 500 feet away. That’s the whole point of safety—when something goes wrong, you fix it, you don’t ignore it. We don’t pretend our AVs are perfect, but we work every day to make them safer than the humans they’re replacing.

Another common concern: what about cybersecurity? If these vehicles are connected to the internet, can someone hack them and make them deliver packages to the wrong place, or even cause harm? That’s a real risk, and we take it extremely seriously. Our automated delivery vehicles have multiple layers of cybersecurity: encrypted communication between the vehicle and the control center, firewalls that block unauthorized access, and regular software updates that patch any vulnerabilities. We test our cybersecurity by hiring third-party hackers to try and break into our system—if they find a flaw, we fix it before anyone else can. It’s a cat-and-mouse game, but it’s one we play every single day because our customers’ trust (and their packages) depend on it.

Let’s also talk about the future, because safety isn’t a static thing. Right now, our AVs are only allowed on certain streets—residential areas, downtown shopping districts, college campuses. We’re not letting them on the highway anytime soon, because highway speeds add a whole other layer of risk. But as our technology improves, we’ll expand to more areas. We’re also working on better pedestrian detection for night time, and software that can handle construction zones where traffic signs are missing or temporary. Last year, we tested a version of our autonomous delivery van in Phoenix during monsoon season, and it performed way better than we expected—even when roads were flooded, it slowed down, used its radar to check for obstacles, and waited until it was safe to move. That’s the kind of testing that matters, not just closed courses on sunny days.

I get it—change is scary. When you order a pizza, you want it to get to you safe and on time, and the idea of a box on wheels doing that feels new, even weird. But here’s the reality: human delivery drivers are more likely to be in a crash than the AVs we build, more likely to get into a fender bender, more likely to rush through a stop sign to make a delivery by the deadline. We’re not trying to replace human workers—we’re trying to take the most dangerous parts of the job (driving) and make them safer. A lot of the people who work for us are former delivery drivers, now they work in our control centers, or in maintenance, or helping test our vehicles. It’s a transition, not a replacement, and we’re handling it as safely as we handle the tech.

So, how safe are delivery autonomous vehicles? Are they 100% safe? No, and I’ll never say they are. But are they safer than the alternative, which is human-driven delivery vehicles right now? Absolutely. We’re still learning, still testing, still fixing things when they go wrong, and that’s exactly what a safe technology does. If you’re curious about how our vehicles work, or want to learn more about our safety protocols, you can check out all our resources on the site, from the Autonomous Delivery Car page to the Automated Delivery Vehicles overview. We’re not hiding anything—we want people to ask questions, to test us, to hold us accountable, because that’s how we get better.

If you’re a business looking to partner with us for delivery, we’d love to talk. We’re not here to sell you a fancy buzzword—we’re here to give you a safe, reliable way to get your products to your customers, every single time. The future of delivery is autonomous, and the safest way to get there is by taking it one step at a time, listening to what people need, and building technology that works for everyone.

References:
National Highway Traffic Safety Administration (NHTSA). (2022). Critical Factors in Traffic Crashes.
Insurance Institute for Highway Safety (IIHS). (2023). Safety Performance of Low-Speed Automated Delivery Vehicles.
University of Michigan Transportation Research Institute (UMTRI). (2023). Real-World Safety Outcomes of Autonomous Delivery Fleets.
Federal Motor Carrier Safety Administration (FMCSA). (2022). Cybersecurity Guidelines for Automated Commercial Vehicles.

Send Inquiry
Send Inquiry