Who’s minding the Robotaxi Store?

If anyone believes the robotaxi hype cycle has passed from its “peak of inflated expectations” (with GM shutting down Cruise) to a “trough of disillusionment,” think again.

In 2025, robotaxis are setting up as a hot ticket agenda among naïve investors and a fresh crop of newshounds.

Robotaxi hype, after all, will be “breaking news” for all those eager influencers and cub reporters who hadn’t before covered the topic, e.g., the self-driving Uber that killed a woman in Arizona in 2018. It will also herald a reprise for automakers who still yearn for a future of “mobility as a service.”

Helping to put autonomous vehicles back on the hype cycle are:

• Tesla’s Cybercab

• A torrent of AI advancement propaganda, e.g., Tesla’s so-called “end-to-end” trained AI stack; Waabi’s fully driverless technology built on GenAI

• Political power already gained by Elon Musk in the incoming administration.

In a nutshell, there is no doubt we will see more robotaxis in 2025 and beyond.

To be clear, robotaxis are proliferating not because some magical technological breakthrough has improved them, nor because Musk has effectively bullied an end to the tough AV regulations that have hampered tech innovations.

Bryant Walker Smith, a University of South Carolina law professor, told us, “Technical, economic, and legal feasibility are distinct concepts … [but] companies can and do deploy Autonomous Vehicles (AVs). They can and will deploy AVs tomorrow.”

Phil Koopman, safety expert and professor at Carnegie Mellon University, went a step further. In a recent podcast, entitled “Five-Step Playbook for Deploying a Cybercab with No Regulatory Approval,” he laid out specific steps for carmakers and tech companies to follow toward robotaxi deployment. Although he said he objects to companies skirting the rules, he explained that there is a “plausible pathway” to faster commercial robotaxi deployment for regulatory scofflaws with a propensity for breaking things, moving fast and asking for forgiveness later.

The scary part of this scenario, explained Koopman, is that “It only takes one company to do this at scale, which will trigger a race to the bottom.”

I recently asked Koopman for real-world examples that illustrate his point.

Step 1:
Koopman’s hypothesis: Build a car with no steering wheel, no gas pedal, no brake pedal, no turn signal. Just a touchscreen in front of the driver.

Real-world example: Building an AV driven only by a game console controller might seem impractical or even unattractive to many people. But this describes Tesla’s Cybercab, which has no steering wheel. A recently surfaced video shows a Cybercab getting loaded for a display at the Petersen Automotive Museum. The Cybercab shown in the video is driven by a video game-type controller wired into the car.

Step 2:
Koopman’s hypothesis: You can meet the Federal Motor Vehicle Safety Standards( FMVSS) specs by self-certifying the robotaxi, with no requirement to show your homework until asked. You buy time to deploy.

Real-world example: In late 2023, Zoox’s robotaxi became the first fully autonomous, all-electric passenger vehicle deployed with FMVSS self-certification. It has no steering wheel or pedals and can carry up to four passengers. Calling it “first of its kind self-certification,” Zoox proudly announced that its vehicles meet “the federal vehicle safety performance requirements needed to operate on public roads.”

Step 3:
Koopman’s hypothesis: Add a self-driving capability that works most of the time. Claim that the person in the car is still the driver because he/she can intervene — in a crisis — with touchscreen driving controls.

Real-world example: This is straight out of Tesla’s Full Self-Driving (FSD) playbook. Marketing this technology as “full self-driving,” Tesla paradoxically caution the human inside to be ready to drive when the Tesla—suddenly—can’t. Tesla — or any robotaxi company — could apply this principle to its robotaxis, requiring the passenger to intervene in a crisis.

Step 4:
Koopman’s hypothesis: Hire remote drivers. When the car suddenly can’t drive itself, the out-of-car safety drivers can take over … maybe.

Real-world example: Meet Vay, now live in Las Vegas, which touts a technology that enables remote “teledriving.”

Trained drivers operate from a teledrive station. They have steering wheel, pedals, and other vehicle controls developed to meet automotive industry standards.

Camera sensors transmit the vehicle’s surroundings to screens at the teledrive station. Microphones send traffic sounds, such as emergency vehicles and other warning signals, to the teledriver’s headphones.

Vay is already operating in Las Vegas, delivering cars to you by “one of Vay’s teledrivers within minutes.” The company says you can drive it anywhere yourself. “End your rental anywhere in Las Vegas. Park yourself or hand the car to one of our teledrivers, who will park it for you.”

Step 5:
Koopman’s hypothesis: Partner with an existing ride-hail network. In states that don’t permit robotaxis, claim that this is just a ride-hail service with human drivers overseeing an autonomous feature.

Mission accomplished: By framing this gimmick as a “Level 2 supervised driving service,” you are already deploying Cybercabs at scale.

One day someone will notice this dodge as bait-and-switch. But Koopman pointed out, “The conclusion will be they are already operating without drivers, so what’s the big deal?” In short, you’ve successfully trained the public to stop questioning the regulatory validity, safety or liability of robotaxis.

At Tesla’s Cybercab announcement event, Musk said that a Cybercab will cost less than $30,000. He also promoted purchases of Cybercabs by consumers in bulk to operate fleets and generate income through robotaxi services. But really, does anyone think you can make money running Cybercabs?

Koopman does. He gives “bonus points for getting private individuals to purchase the cars that are put on the ride-hail network to deflect liability from the manufacturer and transportation network operator.”

The bottom line:
Overruling premarket type approval for new motor vehicles, the U.S. Congress created an extremely permissive U.S. regulatory environment for new vehicles — including robotaxis — rooted in “self-certification.” Moreover, automakers and tech companies have no need to adhere to FMVSS that specifically addresses automating driving. Because no such FMVSS exists.

Sussing out “fully autonomous vehicle” companies’ claims is only getting harder.

Smith, nonetheless, stresses, “a lack of rules ‘specific to’ automated driving does not mean a lack of rules ‘relevant to’ automated driving.”

For safer robotaxi deployments, what must happen in 2025 is not a single-bullet federal regulation change. Rather, authorities — both federal and state — must flex their collective power, taking a hard look at self-driving technology claims and their robotaxi business practices.