Autonomous Vehicles Are a Moral Imperative

Autonomous vehicles are 2025's Trolley Problem.

A trolley is headed down a track, and it will kill 40,000 people every year unless stopped. You could stop the trolley by pulling a lever, but doing so may cost some low-wage jobs and give a moral win to techno-optimists. Do you pull the lever?

For many leaders on both ends of the political spectrum, the answer is a resounding "No." Driverless cars are here, and politicians are not happy about it.

Earlier this summer, Boston city officials demanded Waymo slow-roll its entry into their city. Councilmembers joined unions and activists in opposing driverless cars, even taking umbrage at Waymo making a map of Boston without a community input process. And opposition isn't just confined to the reflexively anti-tech left; Missouri Republican Senator Josh Hawley recently announced a plan to introduce federal legislation banning driverless vehicles nationwide.

“This is too big, and too important, for us to be reactive,” said Boston ‘Chief of Streets’ Jascha Franklin-Hodge in opposition to autonomous vehicles.

And I would agree, but in the other direction.

Almost 1.2 million people across the world die every year in car crashes and up to 50 million are injured, including more two million Americans. And crash victims often die early in their lives, making car crashes the single largest cause of life-years lost globally.

And technology is on the precipice of largely ending this.

But politicians have come to believe that their only role with respect to new technology is to constrain it, to mitigate its impacts, to monitor it and police it and perhaps prohibit it entirely.

But occasionally a technology comes along that is unequivocally a moral good, one that is hard to debate on the merits in any reasonable ethical framework. As the data comes in showing remarkable improvements in safety versus human drivers, autonomous vehicles appear to be exactly that. In the face of mounting safety data, the politicians who demand we slow-walk driverless cars are no better than the vaccine skeptics currently running HHS, sacrificing an unambiguously good technology out of a reflexive fear of things they consider unnatural.

What We Know

I don't fault the average person for not knowing that autonomous vehicles are upon us. The average New Yorker or Bostonian – let alone the modal European – views driverless cars as something far out on the technical horizon, perhaps on par with fusion power or quantum computing.

This is partly due to autonomous vehicles' disappointing hype cycle almost a decade ago. WIRED declared 2016 "the year autonomous vehicles took the wheel," a proclamation that was more about Tesla and Uber's overzealous marketing efforts than the reality of the technology on the ground at the time. As reality set in that full autonomy wouldn't be coming that year – or the next – many journalists and observers wrote the concept off entirely.

But for people in Los Angeles, San Francisco, and Phoenix, autonomous vehicles have been the reality for the past few years. And the safety data is very, very promising.

In particular, Waymo's safety track record shows consistently strong performance compared to human drivers. According to Swiss Re, over 25 million miles driven, Waymo saw 92% fewer bodily-injury claims and 88% fewer property damage claims than human-driven vehicles. And a peer-reviewed study covering 56.7 million rider-only miles through January 2025 showed significant reductions across key crash categories:

• 92 % reduction in pedestrian injuries,
• 82 % reduction in cyclist and motorcyclist injuries, and
• 96 % fewer vehicle-to-vehicle crashes at intersections

... and of the crashes that happened, almost all were caused by human-piloted cars crashing into driverless vehicles.

Of course there are caveats; Waymos generally avoid freeway driving and benefitted from good, predictable weather conditions in Waymo's early markets. But even if these numbers are off by a factor of 2-5x, they're still transformative and represent tens of thousands of lives saved every year. It's unambiguously the accidental death equivalent of a cure for cancer.

But for the media, the fear of being burned by another driverless car hype cycle has meant that AVs are one of the most under-covered technology stories in the US today, drawing remarkably few headlines for a technology on the cusp of reworking how we live and preventing thousands of unnecessary deaths per year.

Three years ago, there was still good reason to be skeptical about driverless vehicles' ability to adapt to different driving conditions and scenarios. Waymo's initial markets offered high-quality roads and sunny, warm weather, and Tesla's FSD – which is a Level 2 system often erroneously described and used as Level 4 like Waymo – suffered from several high-profile crashes.

But self-driving technology has made strides against those objections over the past 24 months. Since last year, Waymo has been conducting winter testing in Buffalo and actively expanding in Austin and Atlanta—hardly winter wonderlands but not 72 and sunny year-round either. While the cars made headlines for pulling over in San Francisco fog in 2023, the driving condition was arguably just as unsafe for human drivers. And Waymo's safety record stands in contrast to Tesla's mixed history, although the latter has been far more aggressive with pushing the envelope and putting bleeding edge technology in consumers' hands – which might be a good thing, as we'll discuss later on.

Regardless, the technology will only get better over time as training data accumulates and driverless cars encounter more edge cases. While true Level 5 autonomy (any time, any road, any conditions) is likely still a decade out, it is more likely than not that Level 4 geo-fenced autonomy can solve 95%+ of car trips within 2-3 years. That's a lot fewer crashes.

Of course, politicians and activists who recognize driverless vehicles' safety advantages still have objections, which generally fall into one of three buckets:

1. "It may work elsewhere, but it won't work here"

This is the kind of exceptionalism in which New York and Boston specialize – the same navel-gazing reason why these cities can't adopt transit policies that have worked throughout Europe and Asia. Everyone claims their city is unique and different as a way to resist change. Case in point:

Sixty-one pedestrians were killed on NYC streets last year; it's likely the vast majority of them would not have been slain by autonomous vehicles.

2. "People should take transit instead"

Look, I love mass transit. I take the NYC subway almost every day. I've argued for mass transit funding and policies that encourage transit like NYC's congestion pricing, a genuine success story since its rollout earlier this year.

But the vast majority of the country's land use does not lend well to mass transit. Most of America is too sparse and too decentralized for mass transit to be effective even if we could snap our fingers and remove the very real barriers to building it at any kind of scale. And that's before we get to the country's cultural aversion to mass transit – a queasiness amplified by many politicians' eagerness to treat transit as a de facto homeless shelter.

Even in urban centers, there are plenty of situations where a car is the better tool. I happily take the train to the airport when I'm traveling by myself, but I'm not schlepping three young kids and luggage to JFK on the subway. There's simply no realistic way to divorce our country from the car; pretending otherwise is simply advocating for the status quo of injuries and deaths.

3. "Jobs!"

Autonomous vehicles will undoubtedly eliminate a lot of jobs. The 3.5 million full-time truck drivers and 500,000 taxi and limo drivers – as well as the estimated two million Uber and Lyft drivers – are in jeopardy. There's a real human impact here.

But responding to new technology by saving obsolete jobs is the wrong response. In a nation with 4.3% unemployment, assistance and retraining programs are the right place to start – anything else is simply kicking the can down the road while perpetuating the epidemic of traffic deaths. American labor unions would be wise to look toward their European counterparts' embrace of technology and push for training and direct assistance rather than encasing obsolete jobs in amber.

To a Policy Framework

At the moment, the battle lines are simply whether we should allow driverless cars to proceed – that is, whether Waymo and its competitors should be permitted to operate on city streets.

But given the human cost of not having this technology, we should be asking how we can take steps to roll out the technology faster. If accelerating driverless cars by 10 years would save 400,000 lives, why not launch an autonomous vehicle equivalent of Operation Warp Speed – a government-led initiative to remove regulatory blockers and leverage the State's production capacity – to get driverless cars on streets faster? Are deaths from a pandemic fundamentally worse than deaths from car crashes?

Rather than simply looking to legalize driverless vehicles, government should think hard about how the state could use its power to actively encourage autonomous vehicle development and commercialization. A few ideas:

1. Data as a Public Good: Safety data-sharing could mirror the FAA’s Aviation Safety Information Analysis and Sharing (ASIAS) program, credited with reducing U.S. airline fatalities by 83% since the 1990s. A federal AV safety clearinghouse, fed by anonymized operator data, would let regulators and researchers benchmark performance across companies and circumvent bad faith accusations of data manipulation or cherry-picking.
2. Regulatory Sandboxes: Governments can create urban testbeds where AV firms operate under modified rules while sharing safety data with regulators. Singapore’s Land Transport Authority, for instance, set up AV testing zones in 2017 and has since approved over 10 companies for pilot operations. In the U.S., only a few states (e.g., Arizona, California) have vaguely comparable frameworks. Formalizing sandboxes nationally could accelerate safe testing while maintaining oversight.
3. Fleet Adoption and Procurement: Public agencies can play a unique role as anchor customers. For instance, the U.S. Postal Service already operates over 225,000 vehicles—the world’s largest civilian fleet—making it an ideal candidate for AV pilots in last-mile delivery. Similarly, autonomous shuttles could supplement public transit, as seen in Jacksonville, FL, where pilot AV shuttles moved over 100,000 riders. Government procurement sends strong market signals while providing controlled environments for scaling.
4. Liability & Insurance Innovation: Liability uncertainty is one of the biggest barriers to AV adoption. While NHTSA data show that 94% of crashes involve human error, insurers struggle to price AV risk. Governments could create federal reinsurance pools similar to terrorism risk insurance (TRIA), enabling carriers to underwrite AVs with reduced uncertainty. Standardized crash-reporting frameworks, tied to black-box data, would further build insurer and public trust.
5. Incentives and Funding: Sometimes the direct approach is the most impactful: government paying for what it wants to see. Tax credits, accelerated depreciation, and R&D funding could mirror clean energy policy successes. For example, the U.S. federal EV tax credit of up to $7,500 spurred adoption of over 3.5 million EVs. A similar incentive structure for AV fleets or AI R&D could de-risk early capital outlays and attract more entrants.

While I've previously criticized Tesla's aggressiveness bringing not-fully-baked technology to market, there's an ethics argument for taking some risk with driverless technology. The baseline, after all, isn't no harm whatever – it's 40,000 deaths per year. As long as the alternative is better than that, it makes rational sense to roll it out.

Of course, humans aren't fully rational creatures, and situations where we feel in control – such as piloting our own car – are far less scary than those in which we do not. While commercial air travel is incredibly safe, passengers' lack of control strikes fear in many flyers. So government interventions such as limiting AV operators' liability for crashes – modeled after similar policies that successfully enabled the development of many vaccines – is probably a step too far, although it would surely accelerate AV adoption.

Of course, even orders of magnitude greater adoption of autonomous vehicles won't completely solve the problem of road deaths overnight. Many people will still want to drive their own cars, and a number of fatal crashes occur in situations unlikely to be replaced by driverless cars – illegal street racing, for instance – without restrictions on human driving.

To that end, I wouldn't be surprised to see the ethics of human-piloted cars grow as a question in the back half of the 2030s into the 2040s, as the death toll from human error becomes increasingly hard to justify. I'd expect some European and East Asian societies – that place a higher priority on the collective good over individual rights – to restrict the rights of people to drive their own cars. (The US is currently rolling back childhood vaccine mandates, so I don't expect similar policies here any time soon.)

Driverless cars are coming. While tech-averse officials may push back, it's going to be increasingly difficult for cities like Boston and New York to maintain their firewalls against AVs as the safety data gets clearer, the technology gets cheaper, and more consumers experience autonomous vehicles when traveling to other cities and wonder "Why can't we have this at home?"

Rather than fighting against the wave, officials would be wise to ensure they're sitting on the right side of history as technology eliminates 40,000 road deaths per year over the next ten years. Not all technology is the same; this is less like Uber or Airbnb and more like the polio vaccine.

Given the lives at stake, do you, elected official, want to be remembered as the one who fought against it?

-Brad Hargreaves