Forgot your password?
Close
typodupeerror
AI Transportation

Waymo Recalls About 3,900 Robotaxis After Some Drove Into 'Freeway Construction Zones' (cnbc.com) 59

Posted by EditorDavid from the danger-zones dept.
CNBC reports: Waymo is recalling almost 3,900 robotaxis in the U.S. to fix software issues after some cars drove into freeway construction zones, according to notices filed with the National Highway Traffic Safety Administration. The voluntary recall, the Alphabet-owned company's second in just over a month, followed 13 known incidents where Waymo robotaxis drove into construction zones on freeways in Phoenix, or entered freeway lanes with active construction in the San Francisco area, the filings published Thursday said... A letter posted to the regulator's website... noted that, "Driving through a closed construction zone increases the risk of a crash..."

[Waymo said in a statement emailed to CNBC] "We voluntarily restricted freeway operations last month while making improvements, proactively notified state and federal regulators, and decided to file a voluntary software recall with NHTSA. We continue to safely serve riders on surface streets in all the cities where we operate...."

The company implemented another voluntary recall in May after some of its robotaxis had driven into flooded zones or standing water. The NHTSA Safety Board also initiated a probe of Waymo after a January incident in which a robotaxi illegally passed a stopped school bus.
This discussion has been archived. No new comments can be posted.

Waymo Recalls About 3,900 Robotaxis After Some Drove Into 'Freeway Construction Zones' More

Waymo Recalls About 3,900 Robotaxis After Some Drove Into 'Freeway Construction Zones'

Comments Filter:

  • Oops! (Score:3)

    by oldgraybeard ( 2939809 ) on Saturday June 20, 2026 @02:36PM (#66201906)
    Time to add another elsif!

    • Seems to me the cars have no way of knowing about works that they cant look up on a database and their visual systems/models arnt yet up to the job of being able to spot possibly badly laid out dividers or cones up ahead, at least until it's too late. Probably not an easy fix.

    • Okay, I think you deserve the Funny mod but I also think it was a weak FP.

      Yeah, my Subject is worse, but... The thing that is going wrong is that we are all part of a mad experiment. Some of the people doing the experiment do have good intentions, but the Waymo robotaxi that tries to follow that road... Well, you know where that road goes.

      But it's a much bigger problem that the humans controlling the various flavors of the experiment have only one intention: MORE MONEY. They already have more money than any

      • It deserves informative.

        The outer constraint system is human coded and fragile, trying and failing to make sure the system fails safe whenever the AI finds a new corner case to fuck up in. Predictably failing especially hard on freeways where you can't be too conservative. Pulling over and phoning home for remote control is no longer an acceptable fail safe.

      • I question why there weren't safety drivers behind the wheel in these cars. The Waymo team should have known that the cars needed to handle construction zones (it's not an edge case, it's a main use case).

        I'm fine with live-testing these cars in production, but there need to be some safety guard rails.
  • Recalls are for when the car has to be voluntarily brought in for an update. Waymo can just push the update. What's the point of a recall?

  • who will do hard time hitting a worker can be char (Score:4, Interesting)

    by Joe_Dragon ( 2206452 ) on Saturday June 20, 2026 @03:08PM (#66201934)

    who will do hard time hitting a worker can be charged as a felony under the state’s “endangerment of a highway worker” or “aggravated endangerment of a highway worker”

    • Probably the guy who makes minimum wage plugging in the chargers.

    • who will do hard time hitting a worker can be charged as a felony under the state’s “endangerment of a highway worker” or “aggravated endangerment of a highway worker”

      You're treating the current law as a standard handed down from on high, incontrovertible and guaranteed-correct, which must be applied verbatim. And, indeed, laws must be applied as written... but that doesn't mean the laws are perfect forever. Laws are written within a context, and when the context changes, the laws have to change.

      In a world where all cars are driven by humans, if you want to protect highway workers one way to do it is to attach serious prison time to killing one and to publicize that f

      • Re: (Score:2)

        by djinn6 ( 1868030 )

        You don't need new laws. Machinery occasionally kill people when they malfunction, including construction equipment. A self-driving vehicle is just another machine. The maker could be considered negligent if they knew this could happen, or it already happened multiple times and no action was taken to improve the software.

        My guess is that in however millions of miles that they've driven, they must've encountered road construction. They might even have a bug filed to address it. This will be revealed in a cou

        • In particular, "road construction areas" are not an unusual edge case. They are a main, important use case.

          So the question is, why didn't Waymo prepare their cars for this important use case? If they knew their cars were not ready for this (or should have known) and let them on the freeway anyway, they are negligent.

    • charged as a felony

      Since when is a felony associated with hard crime? I mean I know one guy who has 34 felony convictions and has never seen the inside of a jail cell.

  • are miracles of bizarchitecture and non-conformance. And often rely on the courts sorting out all possible problems something an automated car may not rely on.

  • Recall to base via construction zones? (Score:4, Funny)

    by 4wdloop ( 1031398 ) on Saturday June 20, 2026 @04:37PM (#66202014)

    So some will crash in construction zones in the process of recalling them?

  • The standard pro self-driving argument (Score:5, Interesting)

    by ZombieCatInABox ( 5665338 ) on Saturday June 20, 2026 @04:41PM (#66202018)

    Whenever self-driving cars are criticized, the standard argument served by the defenders is almost always "Yeah but self-driving cars today already drive better than the average human driver", which, to a certain point, might very well be true.

    But this argument falls flat under scrutiny. See, like most things concerning humans, the quality of human drivers follows a bell curve; There are a few superb drivers, a few shitty drivers, and most drivers are average. But with self-driving cars, all vehicules drive exactly the same way, since they all have the same software. If one of them zooms past a school bus with its stop signals on, they all do. So, for example, if self-driving cars today drive 10% better than the average driver, this also means that they all drive worse than 40% of human drivers out there.

    To be clear: I'm all in favor of self-driving cars, even though I'm among those who criticize them regularly. I've been dreaming of self driving cars since I was a child, and as I'm getting older, I would hope that self-driving cars would allow me to keep my autonomy as my eyesight is getting weaker and my reflexes slower. What I'm saying is that the current approch for self-driving cars is the wrong approch, and the solution is not more sensors, 5G network everywhere, etc. Furthermore, I considers these vehicules, in their current state to be too dangerous to be on public roads.

    But I'm sure the usual binary-thinking simpletons will simply put me in their little "against" box anyway, just like they do when I criticize the current technology of nuclear reactors, so who am I kidding.

    • I hope I don't end up in a situation where I need a self-driving car, because I enjoy driving way too much. And cheap public transportation is great for regular daily transportation.

      • I don't like city or highway driving. Country roads and off road are my preference. And if I'm commuting, I'd rather catch up on my inbox or read a book than white knuckle down the mountain in heavy traffic.

    • Re:The standard pro self-driving argument (Score:4, Interesting)

      by linuxguy ( 98493 ) on Saturday June 20, 2026 @08:35PM (#66202144) Homepage

      > I would hope that self-driving cars would allow me to keep my autonomy as my eyesight is getting weaker ... I considers these vehicles, in their current state to be too dangerous to be on public roads.

      Roughly 6 millions accidents are reported in the US alone every year. And approximately 43,000 people die on the road every year.

      For us to make progress on this front, we have to put these cars on the road and learn from their mistakes. There is no other good way to develop this tech. If these cars were causing tremendous amount of damage to people and property, then yes, they should not be out there. But, data suggests that they are not doing that. Data also suggests that these cars are getting better and safer with each mile traveled.

      If you have a better, safer alternative for us to develop this much needed tech, please share.
      The

      • For us to make progress on this front, we have to put these cars on the road and learn from their mistakes. There is no other good way to develop this tech

        Dumb argument and wrong.

        There are many ways to develop this tech, for example using safety drivers. I'm sure you can think of other methods that would be effective.

        • Also a test site with a test construction site seems to make sense in this case. Then you could even keep adjusting the road markers in realistic ways rather than hope you have encountered every possibility in real life. But then they seem to be successful at asking for forgiveness so that's probably why they don't do it.

      • Re: (Score:2)

        by mjwx ( 966435 )

        > I would hope that self-driving cars would allow me to keep my autonomy as my eyesight is getting weaker ... I considers these vehicles, in their current state to be too dangerous to be on public roads.

        Roughly 6 millions accidents are reported in the US alone every year. And approximately 43,000 people die on the road every year.

        For us to make progress on this front, we have to put these cars on the road and learn from their mistakes. There is no other good way to develop this tech. If these cars were causing tremendous amount of damage to people and property, then yes, they should not be out there. But, data suggests that they are not doing that. Data also suggests that these cars are getting better and safer with each mile traveled.

        If you have a better, safer alternative for us to develop this much needed tech, please share.
        The

        Yes, train your drivers better and have laws that take the bad ones off the road.

        That's why the US has a road fatality rate of 14 per 100,000 pop and the UK has 2.6 per 100,000 pop. Your western countries average around 5 per 100,000 pop (Canada is 4.7, Australia 4.5, France 4.9, Germany 3.3) and we're only really beaten by the Nordic nations who are insanely safe. Before you whine, the per mile statistics aren't any better and the only reason the US drives more is because you won't walk to the shops.

      • If you have a better, safer alternative for us to develop this much needed tech, please share.

        Closed environments and simulations. Simulations are better in particular because you can create test situations trivially, so you can test on e.g. a thousand variations of the same onramp. You can't really build the vision models in simulation, but that's OK, because you can build them by logging data from cars where the computer isn't controlling anything and therefore isn't endangering anyone.

        This isn't new, though, this is obvious. You just want to move fast and break things.

      • I do not trust the companies doing the programming.

        The company with arguably the best software engineers on the planet are completely incapable of writing a reliable operating system and you expect a company with theoretically less-capable engineers to write software maneuvering a multi-ton vehicle at high speeds around other humans?

        Yeah, no. (the results so far are impressive in many ways, but still; yeah, no.)

    • So, for example, if self-driving cars today drive 10% better than the average driver, this also means that they all drive worse than 40% of human drivers out there.

      And? They still drive 10% better than the average driver. And I realize that number is just an example, not intended to be accurate, but I still feel like I should point out that, statistically, it's too low.

      The fact that the self-driving cars will all concentrate their worst behaviors in the same regions of the space of all driving conditions doesn't change the fact that, on average, they're quite a bit safer than human drivers. This wouldn't be true if the roads somehow changed so that the problemati

      • And? They still drive 10% better than the average driver

        That's the propaganda number. "Better" isn't a scientific unit.

        If you want to make it a scientific number, you need to compare like against like. Same driving times, same driving conditions, same driving speeds, same roads (for example, Waymo avoids tricky intersections). If you haven't done that, then you're not being scientific. But to be scientific, Waymo would have to be transparent and open with their data.

        And Google is the opposite of transparent these days. Don't let anyone know more than they ha

        • Also compare to new cars with modern safety equipment. Because that's essentially what a self driving car is. There is no self driving 80s pickup with no visibility.

        • If you want to make it a scientific number, you need to compare like against like. Same driving times, same driving conditions, same driving speeds, same roads (for example, Waymo avoids tricky intersections)

          Bah. If a human driver increased their safety and reliability by avoiding certain situations, would you call them a worse driver for it?

          Waymo would have to be transparent and open with their data.

          They provide full access to the regulators, and they've allowed academic researchers full access. Putting it all online would be more transparent, but they're a business and they have up and coming competitors.

    • Every human drive does stupid stuff *sometimes.* Even accidentally wander into a construction zone.

    • ut this argument falls flat under scrutiny. [snip] ... But with self-driving cars, all vehicules drive exactly the same way, since they all have the same software.

      Except all those self-driving vehicules [sic] drive far better than the average human. They may not drive as well as the *best* human, but that's just you colossally missing the point.

      Your school bus example is a good one. There's a reason why those signs exist: humans zooming past those signs have a tendency to turn kids into a red streak smooshed into the bitumen, whereas Waymos have yet to hit a person in any condition.

      Furthermore, I considers these vehicules, in their current state to be too dangerous to be on public roads.

      Rather than dreaming of a future of self driving cars, maybe reflect on a past where y

      • Except all those self-driving vehicules [sic] drive far better than the average human

        We don't really have the data available to determine that information.

      • Except all those self-driving vehicules [sic] drive far better than the average human.

        So, you're basically refuting my refutation of your point by simply repeating your point ?

        but that's just you colossally missing the point.

        Oh, the irony...

        whereas Waymos have yet to hit a person in any condition.

        https://foxbaltimore.com/news/... [foxbaltimore.com]
        https://www.latimes.com/busine... [latimes.com]
        How many waymos are out there right now ? How long have they been on the street ? The only reason waymos haven't killed any school kid so far is pure luck. You would know that if you had payed any attention in statistics class. "Nothing bad has happened in the handfull of times I've done this thing, so nothing bad ever will" is not a val

    • You're in favour until 50 years into the future it becomes the only way to drive, and the robotaxi cartel charge you a fortune every time you want to go somewhere.

    • It's not that simple, claiming they all drive the same way ignore circumstances of reality, driving past a school bus from sun glare blinding a camera doesn't happen if another one is shifted a foot so the side as the angles are different with the exact same programming.
      Meanwhile more humans crash from sun glare regularly.

      Also the claim that if it's slightly better it's still worst than the best, but think of the hundreds of lives improved from not being crippled/killed by humans who aren't the best. That'

  • Passenger: Hey, you just sped pass that Bridge Out sign!

    Johnny Cab Auto-driver: Nah. It'll be fine.

  • Did they cut back on the number of operators? (Score:3, Insightful)

    by rsilvergun ( 571051 ) on Saturday June 20, 2026 @06:24PM (#66202076)
    We found out in a congressional hearing that the dirty Little secret of Google is that their self-driving cars are actually just remote controlled cars that occasionally use some fancy Lane assist features. But when anything needs to be done that's even slightly complicated it's a human being in the Philippines driving the car.

    I don't want the damn things on my road not that it matters. There is so much money involved I don't get a say and neither do you. So I've got remote control cars being piloted by people in foreign countries over high latency internet connections and I just got a kind of accept that that's a thing I need to worry about now.

    • We found out in a congressional hearing that the dirty Little secret of Google is that their self-driving cars are actually just remote controlled cars that occasionally use some fancy Lane assist features.

      No they aren't. "Remote control" is used only in disengage scenarios. This always results in the ride pausing while a person takes over. Go look up how often that happens, and maybe read that transcript again.

      But when anything needs to be done that's even slightly complicated it's a human being in the Philippines driving the car.

      The centre in the Philippines has zero access to drive the car. Zero. None. No remote control function what so ever. That is done from either Arizona or Michigan. Maybe read that transcript again.

      I don't want the damn things on my road not that it matters.

      I don't want idiots on Slashdot, but we can't always get what we want. Case in point: your post.

      • "Pausing" is no longer a safe option on the freeway by complete coincidence.

      • You know you could just look up the Congressional hearings yourself. Were they admitted everything I said.

        It's all on the record. It's funny but Google says different things when they are under oath then when they are talking to you. Almost as if they don't respect you in the slightest...

  • They should pass a driving test, too. (Score:3)

    by devslash0 ( 4203435 ) on Sunday June 21, 2026 @05:18AM (#66202428)

    How is it that when people want to drive they need to pass a test, but robotaxis don't?

    Someone should build a complex robotaxi-specific driving test field. Then, every time a new version of driving software is released, it should be tested and recertified before being allowed on the roads.

  • It's not a recall when all you do is actually just updating the software over the air. If the vehicle is really having to go back to the dealer/garage to have its software physically updated, yeah, THEN you can talk about recall. But hee "recall" just sounds much more sensational..

  • Wham-o Does it Again (Score:3)

    by BrendaEM ( 871664 ) on Sunday June 21, 2026 @01:15PM (#66202892) Homepage
    If you have an AI company, you are above the law, and have no legal responsibility whatsoever, for now.

Slashdot Top Deals

Shortest distance between two jokes = A straight line

Close