Are self-driving cars safer than cars with drivers?

This BBC podcast is supported by ads outside the UK.

Want to stop engine problems before they start?

Pick up a can of C-Foam Motor Treatment.

C-Foam helps engines start easier, run smoother, and last longer.

Trusted by millions every day, C-Foam is safe and easy to use in any engine.

Just pour it in your fuel tank.

Make the proven choice with C-Foam.

Available everywhere.

Automotive products are sold.

Seafoam!

Suffs!

The new musical has made Tony award-winning history on Broadway.

We demand to be home.

Winner, best score.

We demand to be seen.

Winner, best book.

We demand to be quality.

It's a theatrical masterpiece that's thrilling, inspiring, dazzlingly entertaining, and unquestionably the most emotionally stirring musical this season.

Suffs, playing the Orpheum Theater, October 22nd through November 9th.

Tickets at BroadwaySF.com.

Hello, and thanks for downloading the More Or Less podcast.

We're the program that looks at the numbers in the news, life, and on the streets.

And I'm Lizzie McNeil.

Loyal listener Ambaris Japuria got in touch to ask about a post he saw on social media that made some suspicious-looking claims about the safety of cars driven by humans versus cars that drive by themselves.

The post stated that Tesla's cars are 10 times safer than human-driven cars and Waymo's self-driving taxis are five times safer.

Now we've checked and these claims do come from reports published by the companies themselves.

Now I've been driving for a while.

I've come across plenty of really really bad human drivers, but I also know self-driving cars aren't perfect either.

I mean even the Batmobile crashed crashed into Superman in that one film.

Anyway, back to Ambrush's question.

Is the difference in safety between self-driving cars and human-operated cars really that high?

You hear the stories about how human drivers are terrible.

They get drunk.

They don't pay attention.

They get distracted.

That's Mark McCarthy, a senior fellow at the Brookings Institute Center for Technology Innovation.

And then you think, you've got this automatic machine that doesn't get drunk, it doesn't get distracted, it doesn't get confused, but it's a mistaken logic.

I mean, monkeys don't get drunk, but we don't think that they're going to be any better at driving than people are.

Now, I can hear our loyal listeners begin to type.

So yes, some monkey species actually do get purposefully drunk.

We're aware, we support, we don't judge.

But that's by the by.

The point is that people sometimes automatically assume that machines aren't as fallible as humans.

So they're not as prone to error.

Computers still make mistakes.

As many of the safety engineers have pointed out about autonomous cars, they make the kind of mistakes that people would never make.

Right, well, it's probably good to know what we're measuring self-driving cars up against.

How good are humans actually at driving?

In the United States, humans drive a little over 3 trillion miles a year, and there may be 40,000 or so fatalities for a fatality rate of roughly 1 in 100 million.

And so you're really going to have to go to some good degree to get a better safety record than that.

So for every 100 million miles driven, there is one death.

Now these are just fatalities.

And there's an approximate 6 million car accidents each year in the US, which sounds like quite a big number, but considering there's almost 280 million registered vehicles in the US, that's only a rate of 2.15%.

Right, so that's the humans.

Not as bad as you might think.

Enter the machines.

Tesla's had driver-assisted cars for many, many years.

Those are the kind of cars that will drive by themselves in certain circumstances.

They'll do the steering, they'll do the acceleration, they'll do the braking, all by themselves.

But they're not fully autonomous because they require having a human driver behind the wheel ready to take control at any moment.

Self-driving cars cover two different types of vehicle.

Ones like the Tesla need a human co-pilot.

But what about the ones without a human at the wheel?

The fully autonomous ones, the Waymos, have no driver in the car, just passenger only.

In the United States, they exist in four cities, San Francisco, Los Angeles, Phoenix, and Austin.

These are the self-driving taxis by Waymo.

So let's start with them.

The fully autonomous space age vibe self-driving car.

Are they really five times safer than cars driven by humans?

Well, according to Waymo's own reports, conducted by scientists and looked at by journalists, yeah.

Ish.

With one big caveat.

These cars do not drive everywhere.

They have a very limited area in which they operate.

They restrict themselves to the streets and the areas and the speeds and the weather conditions where they know they can drive safely.

They don't go on highways where many of the most dangerous accidents take place.

So when they do the studies, they have to compare their accident rate with human accident rates just in the area that matches the area that they drive in.

That matching is very hard to get right.

The first issue with this statistic is that WEMA cars do not drive everywhere.

The kinds of accidents that have been reported are relatively minor ones.

Some have no injuries at all.

Only two had severe injuries.

The results really say very little about the kind of safety that the public is interested in, namely fatalities and serious injuries on the road generally, not just in a limited area in a few select cities.

These cars also haven't driven as much as human drivers.

Waymo cars have driven about 70 million miles, but that's a fraction of what human drivers drive.

As I said, human drivers in the United States drive over 3 trillion miles a year.

So Waymo would have to drive hundreds of millions of miles before it could generate statistically reliable data on serious injuries and fatalities.

This means that Waymo have a long way to go before they can prove they match, or ideally better, that one death per million mile rate that human drivers have.

We contacted Waymo and they told us that during the 70 million miles their cars have driven, they were only involved in 48 accidents.

Most of these were minor, but two did result in fatalities.

Also, this figure only counts if you acknowledge the caveat that these cars were driven in specific areas of four cities during safe weather conditions.

Not most humans driving reality.

So what about when you add a human into the equation?

Do you end up with the perfect match?

Well, according to Tesla's Q2 2025 report, Teslos with autopilot engage were 10 times safer than human-driven cars, making them twice as safe as Waymo cars.

But Mark isn't convinced.

The idea that Tesla cars are safer than Waymo cars is totally bogus at this point.

Okay, he's really not convinced.

Tesla cars can drive by themselves, they can steer and they can brake and accelerate.

But there's a human behind the wheel at all times.

And if something begins to go wrong, they can correct it.

So you can't measure their accident rate and compare that to the accident rate of Waymo where there's no one there to take control if something goes wrong.

So any statement that Tesla cars are safer than Waymo is just bogus, period.

Full stop.

As Mark points out, you cannot pit the two cars against each other as you're not comparing like for like.

So while yes, Tesla on autopilot do have good safety stats, it's difficult to disentangle whether the incident avoidance was more to do with the human or the computer.

There are also other issues with this data.

For example, they're comparing a global fleet of Teslas with cars being driven in the US.

So, some of these Teslas are being driven in countries like Norway with really good driving safety records.

And number two, the miles are different.

The majority of autopilot miles will be on a highway.

Tesla's own safety manual states that autopilot should not be used on city streets or on roads where traffic conditions are constantly changing.

Which sounds like exactly the kind of roads and conditions where most accidents happen.

Tesla don't make this claim about their self-driving robo-taxis, and there are currently no private vehicles that do not require a human in the driver's seat.

There are also strict rules about where and when a driver can activate self-drive mode, largely because the self-drive technology isn't good at tackling unknowns.

The fundamental problem that they're dealing with is they have a machine learning system that recognizes the world around them and they train this machine learning system on thousands and thousands of hours of videos and actual driving conditions, right?

But the real conditions that cause accidents are extraordinarily rare.

It's unlikely to be represented in the training data.

Thanks to Mark McCarthy.

And that's all we have time for this week.

If you've seen a number in the news you want us to take a look at, email the program at moreorless at bbc.co.uk.

Until next time, goodbye

from Australia to San Francisco.

Cullen jewelry brings timeless craftsmanship and modern lab-grown diamond engagement rings to the U.S.

Explore Solitaire, trilogy, halo, and bezel settings, or design a custom ring that tells your love story.

With expert guidance, a lifetime warranty, and a talented team of in-house jewelers behind every piece, your perfect ring is made with meaning.

Visit our new Union Street showroom or explore the range at cullenjewelry.com.

Your ring, your way.

With the Wealth Front Cash Account, you can earn 4% annual percentage yield from partner banks on your cash until you're ready to invest.

The cash account grows your money with no account maintenance fees and free instant withdrawals whenever you need it.

Money works better here.

Go to Wealthfront.com to start saving and investing today.

Cash account offered by Wealthfront Brokerage LLC member FENRA SIPC.

Wealthfront is not a bank.

The APY on cash deposits as of December 27, 2024, is representative, subject to change, and requires no minimum.

Funds in the cash account are swept to partner banks where they earn the variable APY.