Unregulated Cybercabs In Five Easy Steps

Gaming the regulatory system has never been easier -- as long as your technology (mostly) works

Pulling out steering wheels does not require regulatory permission, contrary to what you might have heard in the news. Any company that wants to deploy a robotaxi at scale does not need FMVSS exemptions or anyone’s permission. They can just go for it, as long as they don’t mind moving fast and breaking things.(*)

For those whinging about regulators stifling robotaxi innovation, that’s just utter nonsense. There is nothing keeping anyone who really wants to from putting a cybercab or the like out on US public roads coast to coast — so long as their computer driver can go at least a while without an at-fault fatal crash.

Here is a 5-step playbook for deploying a cybercab with no US regulatory approval. It helps if you have a high risk tolerance and aren’t concerned about trampling the gray areas of regulatory oversight in the name of aspirationally saving lives:

(1) Remove the physical steering wheel, brake pedal, accelerator pedal, turn signal stalks, dashboard, etc. Replace them with a touchscreen which displays steering & speed controls and a big red button icon for emergency shutdown. Put it in front of one of the passenger seats. Call those the driver controls. The other indications are probably already on the touchscreen anyway — this is simply a continuation of that trend. Look Ma: No steering wheel! (Just a picture of a steering wheel. That actually does steer the car. It’s the Treachery of Images with another meta-level!)

(2) Self-certify that this vehicle passes FMVSS (Federal Motor Vehicle Safety Standards) by putting the relevant sticker on the car. That’s it — you don’t need permission now, just possible forgiveness later. If asked by press or NHTSA, recycle the quote from Zoox who has already done this: "because there aren’t traditional manual controls, in some cases, that requires, you know, interpreting them in a way that is relevant to a robotaxi." Source: https://www.theverge.com/2024/11/2/24285399/amazon-zoox-robotaxi-nhtsa-fmvss-comply If NHTSA doesn't like it, let them investigate. With competent lawyers (and/or White House intervention) that lets you be on the road indefinitely pending an investigation outcome. And buys you time to build a solid exemption request while the investigation is proceeding as a backup plan. Note that at this point your vehicle is still just an oddly designed regular vehicle, not a robotaxi.

(3) Add a self-driving capability that works most of the time. The rest of the time the occupant can use the touchscreen driver controls. Claim that his means the person in the car is still the driver because they can intervene if needed with the touchscreen driving controls. Tell state & federal regulators this is an SAE Level 2 vehicle because there is a driver supervising the automation, and therefore is exempt from all self-driving regulatory approvals. Bonus points for eliminating the current NHTSA requirement to report SAE Level 2 crash data.

(4) Hire a staff of remote drivers. At first one screen per driver. Then two. Then 100. Ultimately each remote driver has a display with 100 tiny forward camera images and a picture of a red button next to each one that they can tap if there is a problem, making them remote safety drivers. Perhaps located in a cheap labor country. As a practical matter the car is supposed to ask for help when it needs it, so there will be a call center to handle cars that actively need intervention one-on-one with full-screen images. Once this is operational there is no need for a licensed driver in the vehicle. However, the manufacturer can say this is still SAE Level 2 because there is a (remote) driver to intervene by completing the Object and Event Detection and Response subtask (per SAE J3016) when required using a remote copy of the in-car touchscreen controls. At this point you can turn the in-vehicle display into an entertainment display and — presto — you have an SAE Level 2 vehicle (with remote driver) which is indistinguishable to customers in practice from some vehicles today that are seen as legitimate robotaxis.

(5) Deploy on an existing ride hail network. Tell anyone who complains about lack of robotaxi permitting in states that even bother that this is just a ride hail service with human drivers overseeing a self-driving feature. Really no different than current ride hail drivers who can let that same self-driving computer drive them around all day. If you really need to, put a real person in each vehicle to babysit the touchscreen controls until the news cycles settle — and then gradually remove them without attracting attention. One day someone will notice, but the conclusion will be they are already operating without drivers, so what’s the big deal? 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.

Sure, there might be some niggling state laws that get in the way, but that is what lawyers are for — AmIRight? If you have enough billions of dollars to get the technology mostly working, legal battles are no big deal. Fines for ticking off regulators are chump change compared to the burn rate of deploying this technology. Same with settling with the (you hope) small number of crash victims that can't be blamed on someone else. If all else fails shout “China is coming for our economy!!!” to distract everyone. Because right now China requires every robotaxi to have a remote human driver. With a steering wheel. So this is more or less the playbook they are already using — except with regulatory encouragement.

Any company who performs the above steps has put a revenue cybercab onto public roads without needing any robotaxi-specific regulatory approvals. Sure, they might have to fight some court battles and fend off investigations. And maybe the recall bogeyman will come to stifle their innovation (the horror!). But we all have been told that recalls are really just software updates these days, so what’s the problem? And harm to road users no longer counts if that was an old software version instead of the new one. Meanwhile, the company can say they have a robotaxi out there already on the roads without having really broken any rules (just bent some). Anyone who says otherwise must be a Luddite, at least according to the innovators executing this plan or one like it.

To be clear, I do not endorse this path. But this path seems feasible enough we should expect one or more companies to try to do this or something similar in the next few years. Not all companies. But it only takes one, especially if they deploy at scale. And when it happens it could easily trigger a race to the bottom with other companies.

When any such company thinks their self-driving technology is close enough that they can tolerate the risk of deploying at scale, current incentive and regulatory structures will ensure the above will happen. And keep in mind: the risk that companies can tolerate chasing a trillion dollar valuation is almost certainly higher than the risk that vulnerable road users want to have to tolerate.

You can hear a podcast where I walk through these steps on YouTube here:
Article: https://ojoyoshidareport.com/the-wild-west-of-cybercabs-in-five-easy-steps/
Direct video link: https://bit.ly/4iLpKwQ

---

Update 3/1/2025: I have a newer piece that explains why all remote assistants should actually be considered drivers in some sense. Based on that insight, an alternate approach to claiming Level 2 could be to deploy with on-demand remote assistants the same as Level 4 robotaxi companies do. They just provide driving advice when the robotaxi phones home to get advice, and do not continuously watch a camera from the robotaxi or otherwise do continuous real-time supervision. However, because the robotaxi asks for help from a remote assistant to interpret object types or plan a response, that means a human is completing the Object and Event Detection and Response (OEDR) subtask at least some of the time during driving. If a person “completes” the OEDR, that makes the feature Level 2 per SAE J3016 with a remote driver — or at least that is what someone would claim if trying this tactic.

---

(*) This piece is satire, although mostly based on a factual representation of the sad state of the current US regulatory regime and at least some market participants. A large part of how we got here was that in the past companies were somewhat well behaved, and regulations were not needed as a practical matter to enforce red lines that were not crossed in practice. Abuse of the SAE Level 2 regulatory loophole was tolerated to the point it has now been normalized by the industry. This is simply the next step in that entrenched practice. The point of this essay is to show that our system has inadequate checks and balances to protect public safety against a company that is inclined to deploy first and ask forgiveness later. We should fix the system, not encourage companies to game it in this or a similar manner.