The bill that would turn Washington’s streets over to robotaxis runs 24 pages and 542 numbered lines. Getting a permit costs $6 million, a $1 million application fee stacked on top of a $5 million permit fee. Every mile a permitted robotaxi drives carries a 15-cent tax. Crashes reach regulators within eight hours. Operators file quarterly data on how many miles their cars drove empty, how many times one shut down in traffic, and how long each dead vehicle sat there before a tow truck reached it. No company may force a rider into arbitration as a condition of getting a ride.
The word “wheelchair” does not appear anywhere in it.
Neither does “ADA,” “paratransit,” “mobility device,” or “accessible vehicle.” I read the whole thing looking for them. The Autonomous Vehicle Deployment Authorization Amendment Act of 2026, introduced May 1 by Ward 6 Councilmember Charles Allen, contains exactly one disability provision, at lines 251 through 253. It requires an app.
“Provide a user interface, available through a browser-based or mobile device application, that enables individuals with disabilities to request and receive transportation services.”
A rider in a power wheelchair can use that app. What arrives at the curb will not fit the chair.
Key Takeaways
- DC’s robotaxi bill has no wheelchair requirement. Bill 26-0684’s only disability clause mandates an accessible app, not an accessible vehicle. The words “wheelchair,” “ADA,” and “paratransit” appear nowhere in its 542 lines.
- That app clause is a binding command. The network “shall” provide an accessible user interface. The bill asks for software and never asks for a ramp.
- The reporting regime counts everything except accessibility. Quarterly filings track miles by purpose, trips, shutdowns, tow-away times, and crashes. Nothing counts wheelchair trips, wait times, or denials.
- Tesla told the council it is building a wheelchair-accessible robotaxi in Texas, as reported by Wired, with no timeline, no platform named, and no response to requests for comment.
- No operator anywhere offers fully driverless wheelchair-accessible service today. Waymo’s accessible vans are driven by humans, a fact Waymo states in its own rider help pages.
- The real obstacle is securement, not chassis. May Mobility has run an ADA-compliant ramp-equipped Toyota Sienna since 2022 and still needs a human aboard to anchor the chair. Senator Tammy Duckworth’s office found in February that Waymo’s accessible models carry no automated securement system.
- Federal law lets them do this. Under 49 CFR § 37.29, providers of taxi service are not required to buy accessible automobiles. Buy cars instead of vans and the obligation never attaches.
- DC is not a state, so the AV preemption laws that stripped cities in Georgia and Florida of this authority don’t apply. The Council has the power it isn’t using. The written record stays open until July 27.
What the Council Was Willing to Demand
Read Bill 26-0684 next to what its sponsors said about it and a pattern emerges. This is not a timid piece of legislation.
An operator has to log 250,000 testing miles inside the District across at least 180 days before it can hold a commercial permit. Fleets are capped at 200 vehicles until the transportation department signs off on a comprehensive plan. Crashes go to regulators within eight hours, full reports within five business days, and anything filed with federal regulators under NHTSA’s standing general order gets copied to the District within one business day. Quarterly, operators report vehicle miles broken out by whether the car was carrying a passenger, driving to a pickup, or repositioning empty, plus every incident where a vehicle shut down and had to be physically towed away, including how long the removal took.
The liability terms have teeth. A permittee is treated as the licensed driver for traffic law and liability. No company may condition service on a rider agreeing to arbitrate claims against it. And in the section covering registered automated driving systems, if the logs show the software was engaged in the 30 seconds before a crash, the law presumes it was driving and puts the burden on the manufacturer to rebut that by clear and convincing evidence, with tort liability attaching “without the necessity of pleading or proving a product defect.” That section is headed “Private Autonomous Driving System Registration and Liability,” so how far it reaches commercial fleets is a fair question for markup. The drafting nerve is not.
A council willing to presume against the manufacturer and to close off the arbitration clauses every technology platform in America leans on is not a council short on nerve.
It just never wrote down the word.
The bill even creates a fund, seeded by those seven-figure fees and the per-mile tax. After implementation costs, the remainder splits evenly. Half goes to public transit. The other half retrains the rideshare drivers displaced by automation, and the transportation department owes a report on taxi and rideshare displacement within 180 days. The fund compensates the workers automation pushes out. It sets aside nothing for the riders automation leaves behind.
The carve-out worth watching. Section 3a-2(c)(2)(A) makes AV networks follow existing taxi and rideshare law, then exempts “any provision of law that reasonably applies only to an in-person driver.” Much of the District’s for-hire accessibility obligation runs through drivers. Delete the driver and you may delete the duty, without anyone voting to repeal it.
Tesla’s Promise, and Tesla’s Fleet
Into that silence walked a Tesla policy adviser named India Herdman. As reported by Wired, which first covered her testimony, Herdman told council members: “We are in development for a purpose-built, wheelchair-accessible autonomous vehicle.” She described it as an active product being built in Texas.
Wired reported that Tesla did not respond to a request for comment. No timeline was offered. No platform was named. Every outlet that has carried the quote since has credited Wired, and no one appears to have independently transcribed it.
The Washington Times, which had a reporter in the room, reported other Herdman testimony that day. She argued that testing miles logged in other cities should count toward the District’s 250,000-mile threshold, and called the bill’s fees “an unnecessarily burdensome barrier to entry.” So she was there, and she was lobbying on cost.
Set the promise against the hardware. Tesla‘s driverless fleet in Austin is Model Ys, which began carrying passengers with safety monitors aboard in June 2025 and moved toward unsupervised operation over the following year, reaching a metro-wide rollout in June 2026. A Model Y is a crossover. It is not accessible to anyone who cannot get out of their chair.
The Cybercab is the purpose-built one, and Tesla brought it to the National Federation of the Blind convention in Austin on July 3, ten days before the DC hearing. It carries braille lettering on its physical controls, including the interior door releases, plus dedicated space for service animals and external microphones and speakers that announce arrival and accept spoken destinations. The seating sits at wheelchair height.
All of which serves a rider who can transfer out of the chair and into the car. The Cybercab seats two, has no steering wheel and no pedals, and cannot carry a seated wheelchair user. Tesla has never claimed otherwise.
That leaves the Robovan, unveiled in October 2024 with room for 20. It has never received a price or a production date. Speculating that the Robovan is what Herdman meant is speculation, and nothing in any Tesla filing connects them.
Five months before the hearing, Senator Tammy Duckworth’s office stated flatly that Tesla “offers no wheelchair accessible model” and lacks a common conversion platform that would make accessibility upgrades straightforward. That is the baseline the Texas promise is measured against.
Two Kinds of Disabled Rider, Two Opposite Verdicts
The hearing ran past seven hours and heard more than 80 witnesses. Teamsters Local 639, SEIU 32BJ, and ATU Local 689 came to fight the bill over jobs, and that fight got most of the coverage.
The more revealing split was inside the disability testimony, and it is not the split most people would guess.
Claire Stanley, director of advocacy and governmental affairs at the American Council of the Blind, testified in favor. She travels with a guide dog. “I have lost count of the number of times I’ve been refused access because of my guide dog,” she told the council, describing how blind riders who use guide dogs are repeatedly turned away by rideshare drivers. Sandra Neuzil, a legally blind Fairfax County resident, said an on-demand autonomous ride would improve her commute. Conner Cummings of The Arc of Northern Virginia said people like him who can’t drive would get around far more easily.
Follow the mechanism there. Blind riders are excluded by human beings breaking the law. Take away the human and the discrimination has nowhere to live. For them a robotaxi is not a convenience, it is a remedy.
Wheelchair users were in the room too, saying compatible things that point somewhere else. Theo Braddy, executive director of the National Council on Independent Living and a wheelchair user, spoke about what transportation provides: “freedom, dignity, privacy and predictability.” Heidi Case, co-founder of the DMV Disability/Senior Community and also a wheelchair user, made the point the whole bill turns on. As City Cast DC reported her testimony, it has taken several lawsuits and much fighting to win the modest paratransit the District has now.
Nobody sued their way into an accessible taxi by asking nicely. The bill on the table hands out permits without asking at all.
Wheelchair users are excluded by the vehicle. Removing the driver does nothing about the vehicle. Two constituencies, one hearing, one bill, and a technology that liberates one while designing the other out of the picture.
A fact readers deserve. The American Council of the Blind is one of the organizations listed as members of the Waymo Accessibility Network, alongside the National Federation of the Blind and United Spinal Association. Stanley’s account of guide-dog refusals is independently documented and was the basis of a federal lawsuit. Both things are true at once.
Why the Law Lets Them
The regulation doing the work here is 49 CFR § 37.29, titled “Private entities providing taxi service,” and it turns 35 years old this year. Its load-bearing sentence runs eleven words.
“Providers of taxi service are not required to purchase or lease accessible automobiles.”
A taxi company may not refuse a disabled passenger who can use the vehicle it has. If it buys a larger vehicle, a van, that van must be accessible unless equivalent service already exists. What it never has to do is go acquire something other than automobiles in order to put a ramp in the fleet.
Whether a driverless operator counts as a provider of taxi service under a rule written in 1991 is its own unsettled question. Nobody in the industry seems eager to litigate it, because the rule as written is doing them a favor.
The loophole is structural, and it is older than the companies using it. Buy cars, not vans, and you owe nobody a ramp. Uber and Lyft built businesses inside it. Robotaxi operators inherited it intact. Waymo buying Jaguar I-PACEs and Zeekrs, and Tesla building a two-seat Cybercab, is compliant by construction.
Which reframes what Waymo’s policy executive told the council. Matt Walsh said his company has “not been able to identify a platform that is fully wheelchair-accessible while also meeting the unique specifications to retrofit that vehicle with our technology,” again as reported by Wired. That sentence describes a search. Federal law never required the search to succeed. It doesn’t ask anyone to look.
Enforcement is happening, but notice at what. The Justice Department sued Uber in September 2025 under Title III of the ADA, seeking $125 million for riders with service animals and stowable wheelchairs who were refused, charged cleaning fees, or told they couldn’t sit up front. Uber moved to dismiss and lost in March, so the case is now in discovery. Read the theory carefully, though. It is a case about refusing people the vehicle could have carried. It is not a case demanding Uber buy a single ramp, because § 37.29 says it doesn’t have to.
That asymmetry is the entire blind-versus-wheelchair split, written into federal regulation. The law vindicates the guide-dog rider. It has nothing to say to the woman who can’t board.
The Wall Nobody Talks About
Here is where the platform story falls apart, and it is the part missing from the coverage.
May Mobility, working with BraunAbility, put an ADA-compliant rear-entry ramp into a Toyota Sienna Autono-MaaS and phased it across its sites in 2022. It carries one wheelchair user plus two seated passengers, or four without the chair. In-cabin speakers and a display tell riders with hearing or vision impairments when it’s safe to get in and out. This vehicle exists. It has existed for four years.
Something rides along with it, though. May Mobility keeps a human operator aboard its accessible vehicles to deploy the ramp and help. When the company laid out its plan to go fully driverless, it said its wheelchair-accessible vehicles would keep those operators on board even so. Its rider-only service, the one with nobody up front, launched in Sun City, Arizona in December 2023 without a wheelchair-accessible vehicle in it.
May Mobility solved the ramp. It did not solve securement, the business of anchoring a 300-pound power chair and its occupant safely with nobody there to do it. Back in 2023 the company described itself as one of several parties working on an independent wheelchair-docking station and an automated belt-donning system, which is the piece that would let the human operator step off.
Now put Duckworth’s February finding next to Walsh’s July testimony. Her office reported that Waymo has wheelchair-accessible models in its fleet, but that none carry an automated securement system, which is why wheelchair users cannot use Waymo independently. Duckworth asked the company to bring a universal automated wheelchair model to market.
Read Walsh’s sentence again and the two accounts do fit, which is the interesting part. He said Waymo could not find a platform that is wheelchair-accessible while also meeting the unique specifications to retrofit that vehicle with our technology. Waymo’s accessible vans are human-driven and carry none of its autonomy stack. Both statements can be true at once.
What the qualifier does is relocate the problem. It describes a hunt for a chassis, which sounds like a supply issue somebody else has to fix, when the constraint two operators have independently run into is securement. May Mobility has the accessible Sienna and drives it autonomously. It still parks a human inside to work the ramp and the straps.
Securement is a real, hard, unsolved engineering problem, and it deserves more respect than a sentence about platform availability gives it. Anyone can bolt on a ramp. Nobody has shipped a system that safely restrains an occupied wheelchair without hands. That is the honest answer to why these fleets look the way they do, and it is a more interesting answer than the one offered in hearing rooms.
What Waymo Admits in Writing
The most damaging document in this story is not testimony. It’s a help page.
Waymo’s own rider support site says its wheelchair-accessible vehicle service is “provided for people with mobility-related disabilities who cannot ride in our fully-autonomous cars.” Those vans are “driven manually.” By a person.
In San Francisco and Los Angeles you can toggle a WAV on in the app. In Phoenix you cannot; you have to call support and ask them to order one for you. Morris reports that San Francisco’s are run by a third-party contractor. And when Waymo launched in Atlanta, its answer for wheelchair users was a referral to Uber’s accessible service, which does not operate in Georgia.
Waymo publishes no WAV trip counts, no wait times, no fleet size. The company that reports its safety record in injury crashes avoided per week, and whose executive told this council it is “avoiding about 13 injury crashes every single week,” publishes nothing about how long a wheelchair user waits.
Which brings up the quietest failure in the bill. Its quarterly reporting is meticulous about miles, trips, shutdowns, tow-aways, and crashes. It requires none of it broken out by accessible service. No WAV trip count, no WAV wait time, no record of denials. You cannot enforce what nobody counts, and the District chose not to count.
Zoox is worth a line here, because it looks like the answer and isn’t. The vehicle is purpose-built, symmetrical, with wide sliding doors and a cabin the company describes as curb-level. The first-generation vehicle is not wheelchair accessible. A version that is remains in development with no firm date. Kate Holt of the accessibility firm Evinced, who rode one in March, wrote that it leans heavily on sight, with “no auditory prompt indicating where to sit, no spoken instruction explaining how to begin the ride.” Despite the curb-level billing, she wrote, the car “often stopped several feet away from the sidewalk.” A curb-height floor means nothing if the car doesn’t stop at the curb.
And Cruise had a flat-floored, purpose-built Origin that could have carried a chair. General Motors scrapped the Origin in July 2024, taking a $583 million charge and falling back to the Chevrolet Bolt. The accessible variant died with it, five months before GM stopped funding the robotaxi program altogether.
| Operator | Driverless today | Wheelchair accessible | Both at once |
|---|---|---|---|
| Waymo | Yes (I-PACE, Zeekr) | Separate vans, human-driven | No |
| Tesla | Yes (Model Y, Austin) | No model offered | No |
| May Mobility | Yes (rider-only sites) | Yes (BraunAbility Sienna) | No, WAVs keep a human aboard |
| Zoox | Yes | Gen 1 no, WAV in development | No |
The Part That Makes DC Different
The standard version of this story is about preemption, and it’s a good story. State legislatures, often working from industry-drafted text, have declared rideshare and now autonomous vehicles subject to state law alone, stripping cities of any power to impose local operating standards. Those state laws carry no accessibility mandates. Georgia’s autonomous vehicle statute provides that no rule “shall be adopted which limit the authority to operate such vehicles or systems conferred by this Code section.” Missouri’s House passed its version 96 to 48 this year.
The results are on the ground. The accessible-travel writer John Morris, who has tracked this closely, reports that there is no wheelchair accessible Uber or Lyft service in the state of Florida, and none in Ohio. Accessible taxi fleets, which did carry local mandates, got undercut and driven off the road. By Morris’s estimate there are fewer wheelchair taxis on the road today than there were in 2010.
Technology that was supposed to be transformative for disabled people is arriving under laws that guarantee it won’t be.
Except that story doesn’t explain Washington. DC is not a state. No legislature preempted this council. No industry-drafted statute took the pen out of Charles Allen’s hand. The District has exactly the authority Atlanta and Miami lost, and it sat down to write a robotaxi framework with that authority fully intact.
It produced 542 lines, seven-figure fees, a reversed burden of proof, an arbitration ban, ward-by-ward service parity, eight-hour crash notification, and an app.
Oregon shows what the industry asks for when it has to ask. Fighting HB 4085 this February, Disability Rights Oregon told the state House committee it understood that operators, Waymo among them, were pushing for language that would let a network refer a rider to some other carrier when they need a “fixed-frame wheelchair” and no vehicle is available. Executive director Jake Cornett named it precisely: “A referral-only approach creates a two-tier transportation system: most riders receive direct service from the autonomous network, while wheelchair users are told to find transportation somewhere else.”
Cornett’s other line is the one Washington should sit with. “Whether a large autonomous vehicle company ‘can’ serve wheelchair users is a business decision, and one entirely within that company’s control.”
DC’s bill doesn’t even reach the referral standard the industry itself offered in Oregon. It asks for an app.
The record stays open until July 27, and no markup appears to have been scheduled. The council that made operators account for every empty mile and every tow truck, and that asked them to explain how they would keep one ward from waiting longer than another, understands the principle perfectly well. For wheelchair users it wrote one line about an app. The amendment writes itself. Somebody has to write it.
Bottom Line
Washington’s robotaxi bill charges $6 million to enter, taxes every permitted robotaxi mile, blocks arbitration as a condition of service, and demands quarterly data down to how long a dead vehicle blocked the street. It never once says “wheelchair.” Its only disability provision requires an app. Tesla told the council it’s building an accessible robotaxi in Texas, as reported by Wired, with no timeline and no platform named, five months after a senator’s office found Tesla offers no accessible model at all. Waymo’s own help page says its accessible vans are driven by humans, because wheelchair users cannot ride in its autonomous cars. May Mobility has run a ramp-equipped autonomous Sienna since 2022 and still needs a person aboard to anchor the chair. Securement is the real bottleneck, an unsolved engineering problem worth naming plainly rather than describing as a shortage of vehicles. Federal law has never required any of them to solve it. DC, unlike Atlanta or Miami, was never stripped of the power to require it. The council simply didn’t.
Frequently Asked Questions
Are any robotaxis wheelchair accessible right now?
No operator offers fully driverless wheelchair-accessible service anywhere in the United States. Waymo runs wheelchair-accessible vans in San Francisco, Los Angeles, and Phoenix, but its own help pages state they are driven manually by human drivers and are for riders “who cannot ride in our fully-autonomous cars.” May Mobility operates ADA-compliant autonomous Toyota Siennas, but keeps a human operator aboard to deploy the ramp.
Is the Tesla Cybercab wheelchair accessible?
No. The Cybercab is a two-seat vehicle with wheelchair-height seating designed to help a rider transfer out of their chair and into the car. Someone who cannot leave their wheelchair cannot use it. Its accessibility features, including braille controls, service-animal space, and audio arrival announcements, are aimed mainly at blind and low-vision riders.
What did Tesla tell the DC Council?
According to Wired, which first reported the testimony, Tesla policy adviser India Herdman said on July 13, 2026 that “we are in development for a purpose-built, wheelchair-accessible autonomous vehicle” being built in Texas. She gave no timeline and named no platform, and Wired reported that Tesla did not respond to a request for comment. No other outlet appears to have independently confirmed the quote.
Does DC’s robotaxi bill require wheelchair-accessible vehicles?
No. Bill 26-0684’s sole disability provision, at lines 251 to 253, requires operators to provide an app-based user interface that lets people with disabilities request rides. The words “wheelchair,” “ADA,” “paratransit,” and “mobility device” do not appear in the bill’s 542 lines, and there is no fleet-percentage mandate, response-time parity requirement, or accessibility reporting obligation.
Why don’t robotaxi companies just buy wheelchair vans?
Federal regulation doesn’t make them. Under 49 CFR § 37.29, providers of taxi service are not required to buy or lease accessible automobiles, and are not required to acquire vehicles other than automobiles in order to have accessible ones in a fleet. If an operator does buy a van, that van generally must be accessible. Buying cars avoids the obligation entirely, which is the structure Uber and Lyft used and robotaxi operators inherited.
What is wheelchair securement and why does it matter?
Securement is the process of anchoring a wheelchair and restraining its occupant so both stay put in a crash. In a conventional accessible van a driver or attendant straps the chair down by hand. No company has shipped a system that does this automatically, which is why May Mobility keeps human operators on its accessible vehicles. Senator Duckworth’s office reported in February that Waymo’s wheelchair-accessible models carry no automated securement system, meaning wheelchair users cannot use Waymo independently.
Did disability advocates support or oppose the DC bill?
Both, along different lines. Blind and low-vision witnesses testified in favor, because human drivers illegally refuse riders with guide dogs and removing the driver removes that discrimination. Wheelchair users who testified supported the promise while noting the record, including that it took repeated lawsuits to secure the District’s existing paratransit service. The vehicles exclude them regardless of who or what is driving.
Can DC legally require wheelchair-accessible robotaxis?
Yes. Several states, including Georgia, Florida, and Ohio, have passed preemption laws barring cities from imposing operating standards on rideshare or autonomous vehicle companies. DC is not a state and no legislature preempted its council, so the District retains the authority those cities lost. Congress does hold a 30-day review window over DC legislation, but nothing prevented the Council from writing an accessibility mandate.
Is the DOJ lawsuit against Uber about wheelchair access?
Partly, but not about vehicles. The Justice Department sued Uber in September 2025 under ADA Title III, seeking $125 million over drivers refusing riders with service animals and stowable wheelchairs, charging improper cleaning fees, and denying front-seat requests. It targets refusal of riders the vehicles could have carried. It does not require Uber to add wheelchair-accessible vehicles, because federal regulation doesn’t require that.
Can I still weigh in on the DC bill?
The written record for Bill 26-0684 was open until July 27, 2026, and no committee markup had been scheduled as of mid-July. Testimony filed with the Committee on Transportation and the Environment typically posts to the Council’s legislative information system after the record closes.
When would robotaxis start operating in DC?
Under the bill as introduced, on-demand autonomous networks could begin operating no earlier than January 1, 2028. Operators would first need 250,000 miles of testing within the District over at least 180 days, and fleets would be capped at 200 vehicles until the transportation department approves a comprehensive plan.
Michael Kahn
Michael Kahn is the writer, photographer, and publisher behind The Weekly Driver. He cares about how cars drive and what they're like to own. He covers automobile industry news, car shows and events, and new car reviews. The reviews come from behind the wheel: day trips that favor back routes, treating a good meal as half the reason to go. He directs and produces the visual media, matching each car to a setting and mood that fit it. When he's not reviewing new cars, Michael races paddleboards, camels, and ostriches, along with the occasional exotic car on the racetrack, and has driven in every state and country visited.
https://theweeklydriver.com