Robotaxi Test Driver Injuries Mount as Waymo Takes $5B Loan

The people still riding inside the machine

Public arguments about robotaxis usually center on two groups: passengers who pay for rides and people who share the road with the vehicles. A third group gets much less attention. These are the test drivers and safety operators who ride inside autonomous vehicles while the software is still learning, and recent reporting shows they are getting hurt.

Workplace-safety filings show that test drivers for Waymo and Zoox were injured more than two dozen times in 2024 and 2025. The injuries came from the vehicles braking hard, swerving or making other sudden moves.31 Transdev, the contractor that employs and manages Waymo's test drivers, reported 16 injuries at Waymo depots in three cities. Zoox reported up to eight injuries tied to hard braking.31 The workers had sprains, strains and whiplash, and some were out of work for weeks or months.3140

This is not a story about one dramatic crash. It describes a steady, repeated physical cost of training self-driving software, and it is surfacing as the industry's biggest company raises billions to grow faster.

What the injury records describe

The details show a pattern. Transdev reported five Waymo test-driver injuries in 2024 across San Francisco, Los Angeles and Phoenix. One involved hard braking and four involved other erratic vehicle behavior.31 In one case, a car that was parking itself suddenly reversed and spun the steering wheel. The wheel caught the driver's hand and bent it until he heard his wrist crack, and he missed more than two months of work.31

In 2025 the count in those three markets rose to 11, and hard braking caused nearly all of them.31 A Phoenix driver missed 157 days after the vehicle braked sharply with nothing in its path. At least two other reports also say the braking had no apparent cause.31 The only 2025 report that gives a reason is a Los Angeles case where children were playing in the road. That stop cost the worker 175 days off.31 One analysis compared these absences with the eight-day median for U.S. workplace injuries. By that measure, the two longest cases together add up to more than a full worker-year, and in neither case did the vehicle hit anything.37

Zoox's records point to its own recurring problem. Seven of its eight entries specifically mention hard braking, "nogos" or "brake jabs."31 A "nogo" is the company's internal term for a full system shutdown, which often triggers a brake jab.31 Contractors said these jabs can happen several times on one drive, often when the software detects road debris that may not be there.31 One contractor said a nogo at 45 mph throws the operator forward in the middle of traffic, at the exact moment they need to take control.31

The two companies responded differently. Zoox said hard braking is sometimes unavoidable and that the injuries are a small fraction of its millions of test miles. Waymo did not answer questions about the OSHA filings. It said its tens of millions of supervised miles are part of a responsible validation process.31

The recall that didn't end the braking

Zoox's situation is the more pointed of the two. Four of its 2025 injuries happened after it issued a recall to resolve a federal investigation into unexpected braking. Zoox says that recall covered a very narrow behavior and that the later injuries are unrelated.31 The investigation began after motorcyclists rear-ended Zoox vehicles that allegedly braked without warning.2636

Two current and three former Zoox contractors said the modified Toyota Highlander test vehicles were still braking abruptly as recently as July 2026. Three of them said workers were still being injured.3135 Zoox has also reported two crashes to federal regulators in which its purpose-built robotaxi braked for debris and was rear-ended.31 Both companies insist that test-fleet behavior and commercial-fleet behavior are separate issues. Still, one behavior appears in both: braking for things that may not be on the road.

In September, a different kind of worker-safety problem came up. Zoox grounded its Atlanta test fleet after safety drivers reported symptoms they linked to possible gas exposure inside the vehicles. One worker filed an OSHA complaint, and the agency opened an inquiry.38 An OSHA letter said up to 40 workers may have been exposed. Zoox told the agency that 28 employees had ridden in six vehicles with an "unexplained odor," that 10 sought medical evaluation, and that one medical report identified carbon dioxide exposure.38 Some workers' accounts conflict with the company's, and those accounts have not been independently verified.11 One automotive analysis argued the issue has nothing to do with autonomy. It described it as a familiar fleet-upfitting hazard: combustion-engine SUVs loaded with electronics and run for long shifts with several people inside.11 That reading seems right on the technical cause. It also strengthens the larger point that the humans in the test pipeline carry risks the industry's safety messaging rarely mentions.

A dataset that exists by accident

One key detail in the reporting is why these injuries became public at all. Waymo and Zoox show up in OSHA's public data only because they classify themselves as taxi, limousine or transit services. OSHA treats that category as higher-hazard and requires more disclosure.31 Other autonomous-vehicle developers do not appear in the data and may be exempt.3139 Public reporting also applies only to establishments with 100 or more employees. That means injuries in newer, smaller markets may never be visible.31 And since companies don't file annual data until the following year, the 2026 numbers are still unknown.31

The reporting broadly agrees on the counts. Outlets describe more than two dozen injuries, or at least 24.333437 The disagreement is about what the counts mean. One business analysis noted that neither company publishes its test-fleet size or supervised mileage. Without those numbers, there is a raw count but no injury rate. It also noted that 24 injuries over two years would not stand out in conventional transit, which already has a high injury baseline.37 That is a fair caveat. The same analysis then made the observation that matters more: Zoox's purpose-built vehicle has no controls and no employee on board, so its hard braking produces no OSHA entry at all. The industry is moving away from the only dataset that has captured this problem.37

Federal crash data has a similar blind spot. A breakdown of NHTSA's Standing General Order reports counted 1,436 automated-driving incidents from June 2025 through July 2026. Waymo filed 84.3% of them.21 Those reports do not include miles driven, so they cannot show whether risk is going up or down.21 Waymo's own analyses report far fewer injury crashes than human drivers in the same areas. However, independent reviews note that most of that evidence comes from the company itself.25 The published figures also vary by release. One cites 82% fewer injury crashes and 95% fewer serious-injury crashes over more than 270 million miles.13 Another cites 94% fewer serious-injury crashes over 220 million miles through March 2026.36 Neither set of figures says anything about the people sitting in the driver's seat during testing, because the two datasets measure different things.36

Expansion is accelerating anyway

The injury reports have not slowed Waymo. On October 8 it closed a $5 billion term loan, its first debt financing. PIMCO, Blackstone and Sixth Street led the lending group.110 The loan follows a $16 billion equity round earlier in 2026 that valued the company at $126 billion.7 Waymo says the money will speed expansion in the U.S. and abroad. It opened its fifteenth U.S. city in September and has announced plans for Singapore, London and Tokyo.89 This week it also began airport rides in Nashville.6

Even the size of the operation is reported differently. One reference count puts Waymo at 500,000 paid rides a week as of mid-2026.2 A financial report from the same week of the loan gives a figure of about 400,000.3 In either case, entering a new city means mapping and supervised testing first. That means more hours with human operators on board, and the reporting links more test miles to more injuries.3132 Rivals are on a similar path. Zoox added Houston and San Diego in September, bringing it to 12 U.S. locations, and both will start with its retrofitted test fleet.11

Regulation is aimed elsewhere

Current policy debates mostly leave this issue out. California's new SB 1246 deals with first responders. It fines operators whose vehicles block police or firefighters for more than 30 minutes and requires on-site incident technicians. The law does not take effect until July 2028.1518 NHTSA's latest action concerns a petition alleging that Tesla robotaxis sped through school zones. The agency has logged four incidents and no injuries.2427

At the city level, the fight is mostly about jobs. Portland's transportation bureau has proposed fleet caps and a requirement of 50,000 test miles before paid service, at least half of them with a human driver on board.12 Driver unions are pushing for a human operator in every vehicle.12 Minneapolis passed a human-driver requirement, but the mayor vetoed it.2 In Washington, D.C., protesters rallied against legislation that would allow commercial robotaxi service.13

There is an irony here. Local rules that require more supervised miles, or human operators on board, would put more workers in the seat where the injury records show the risk is. None of these proposals, as reported, deals with the occupational hazard itself, such as braking forces, cabin air or injury reporting for smaller operations.

The reading

The evidence does not show that robotaxis are more dangerous than human drivers. Waymo's crash record in its established markets is substantial, even with the caveats.25 What the evidence does show is narrower. The industry's caution, meaning software that brakes hard when unsure, is not free. Contract workers absorb much of that cost, and the main visibility into it comes from a reporting classification that newer operators and driverless vehicles fall outside of.3137 As Waymo uses its new loan to open more cities, the number of supervised test miles will grow. Unless OSHA or state regulators treat test operators as a safety population of their own, the next round of injuries could be less visible than this one.