Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
EU AI Government The Courts

EU Proposes Rules Making It Easier To Sue Drone Makers, AI Systems 66

The European Commission on Wednesday proposed rules making it easier for individuals and companies to sue makers of drones, robots and other products equipped with artificial intelligence software for compensation for harm caused by them. Reuters reports: The AI Liability Directive aims to address the increasing use of AI-enabled products and services and the patchwork of national rules across the 27-country European Union. Under the draft rules, victims can seek compensation for harm to their life, property, health and privacy due to the fault or omission of a provider, developer or user of AI technology, or for discrimination in a recruitment process using AI.

The rules lighten the burden of proof on victims with a "presumption of causality", which means victims only need to show that a manufacturer or user's failure to comply with certain requirements caused the harm and then link this to the AI technology in their lawsuit. Under a "right of access to evidence," victims can ask a court to order companies and suppliers to provide information about high-risk AI systems so that they can identify the liable person and the fault that caused the damage.

The Commission also announced an update to the Product Liability Directive that means manufacturers will be liable for all unsafe products, tangible and intangible, including software and digital services, and also after the products are sold. Users can sue for compensation when software updates render their smart-home products unsafe or when manufacturers fail to fix cybersecurity gaps. Those with unsafe non-EU products will be able to sue the manufacturer's EU representative for compensation. The AI Liability Directive will need to be agreed with EU countries and EU lawmakers before it can become law.
This discussion has been archived. No new comments can be posted.

EU Proposes Rules Making It Easier To Sue Drone Makers, AI Systems

Comments Filter:
  • I get the angst on AI driven cars and deaths, but they have way less accidents than humans. How about we fix things like copy right law, OEMs intentionally bricking or slowing down devices to force upgrades, and the abuses of digital licenses that take away ownership of products. Yes we can do several things at once, but we can't do many things well at the same time.. Priorities...
    • Nah, I'd rather take this at face value. After all, when 100,000 people are harmed and the class-action kicks off, I'll be looking forward to really sticking it to them with my $7 check in five years.

      That'll fuckin' teach them to do it again and again and again.

      Gee, it's almost as if lawyers are helping write this law, 'cause Daddy Esquire needs a new yacht when his new one becomes fashionably old in 3 years. Like a smartphone.

    • by Carewolf ( 581105 ) on Thursday September 29, 2022 @04:41AM (#62923289) Homepage

      No, they dont have less accidents than humans. When they only drive under conditions where humans have 100x fewer accidents, they have only 10x fewer accidents, but then compare that to humans under all conditions to lie and pretend to be safer.

      • by tomkost ( 944194 )
        Hmm, that is an interesting point. I had not considered the pareto of when the AI drives vs human...
      • There are 2/3 more accidents in places where there is snow and ice, but how many automated vehicles are safer in snow and ice?
    • by noodler ( 724788 )

      I get the angst on AI driven cars and deaths, but they have way less accidents than humans.

      I don't think there is any evidence that autonomous cars are safer at the moment. Maybe in the future.

      • Why wait? I don't know how large Tesla's "full self driving" beta is or how often the testers use it versus just autopilot. But It's already pretty rare for me to hit the road and *not* see at least one of those Waymo Jaguar SUVs driving around. And Google is nothing if not good at collecting and analyzing data. The Cruise self-driving cars are very common these days too. It shouldn't be to much of an exercise to run the man vs machine numbers for the two, at least.

        • As long as the new self driving vehicles are compared to a modern vehicle with modern proximity sensors, cameras, and adaptive cruise control rather than comparing a vehicle made in 2022 to some vehicles made in 1985.
    • Right. AI cars only kill people sometimes, so we should focus on the really serious life-threatening stuff like copyright and DRM. Gotcha.

    • Will Russia now get to sue Turkey over those pesky Bayraktars?
  • by registrations_suck ( 1075251 ) on Wednesday September 28, 2022 @08:45PM (#62922805)

    I swear to god, it seems like EU's mission in life is to file lawsuits, create opportunities for filing lawsuits, invent new reasons to file lawsuits, etc.

    I think the entire world could be at peace and harmony with itself...world hunger could be solved, everyone could have a home and plenty of food to eat and everything else they could possibly want, and the EU would be like, "We need to sue somebody!!"

    • IANAL, but I see a basic difference between US and EU legal systems. US courts tend to give almost equal weight to statutes (passed by Congress) and precedents (previous court decisions), while EU courts give much more weight to statutes [wikipedia.org].

      Civil law judges tend to give less weight to judicial precedent, which means that a civil law judge deciding a given case has more freedom to interpret the text of a statute independently (compared to a common law judge in the same circumstances), and therefore less predict

      • Exactly this. The US (and, I think, the UK) have legal systems that make use of previous court decisions. Ultimately, a series of court decisions can wander farther and farther from the law, and create results completely at odds with what the law says.

        In all other EU countries, indeed afaik in most of the world, the law is the law. Previous court rulings are only relevant, insofar as they clarify ambiguities in the law. Even then, the current judge is free to clarify those ambiguities as seems appropriate

        • by sabri ( 584428 )

          The US (and, I think, the UK) have legal systems that make use of previous court decisions. Ultimately, a series of court decisions can wander farther and farther from the law, and create results completely at odds with what the law says.

          Because sometimes, statutory law does not cover all the exceptions that society would really want to a law. In a common law system, judges generally listen to their gut to determine who is at fault and who is not, and then tend to write an opinion around it.

          And of course, you rightfully ask "OH YEAH, GIVE ME AN EXAMPLE".

          Riggs v. Palmer [wikipedia.org] (TL;DR: murderer did not get an inheritance from the person he murdered, despite statutory law saying he was entitled to it).

          • Well, if I were the judge I'd just give him a free pass, and then put in the decision my opinion that there should be a law for cases like that.
            • by sabri ( 584428 )

              Well, if I were the judge I'd just give him a free pass, and then put in the decision my opinion that there should be a law for cases like that.

              Well, that's how the common law system works. The judges opinion becomes part of established law, assuming it is not overruled by a higher court.

              And that's also why I know about it: it was part of my list of cases to brief for law school.

    • by mjwx ( 966435 )

      I swear to god, it seems like EU's mission in life is to file lawsuits, create opportunities for filing lawsuits, invent new reasons to file lawsuits, etc.

      I think the entire world could be at peace and harmony with itself...world hunger could be solved, everyone could have a home and plenty of food to eat and everything else they could possibly want, and the EU would be like, "We need to sue somebody!!"

      Which is why the US is not the capital of frivolous law suits?

      The first rule of Slashdot headlines about the EU is that Slashdot headlines about the EU are always going to be sensationalised, if not completely wrong.

      This move has said that the plaintiff only needs to reasonably demonstrate that the company behind the device/software did not comply with EU regulations. Upon this presumption, the company is then required to turn over evidence that they did, indeed, comply with relevant EU regulations or

    • It seems to me that any copyrighted image used to train a neural net makes that neural net a derivative work, and by the laws of youtube 100% of the monetization revenue should be claimable by the copyright holder of any image/sound/video used in the training set of the AI. DMCA saves the world from Skynet apocalypse!

  • by mkwan ( 2589113 ) on Thursday September 29, 2022 @12:07AM (#62922999)

    About the only tech industry where the EU is competitive these days is automotive, but this legislation would kneecap any hope of a European-made self-driving car. I mean, would you work on a self-driving car project knowing that you, personally, could be sued for a bug?

    It looks like the future of automotive will be American, Japanese, Korean, and Chinese.

    • I dunno. AIs can be copied or cloned easily. If someone has it, everyone has it in a few months. It's not like chip production technology, that's hard to replicate. AI is also getting easier to use, now you can do many tasks with 5 lines of Python, or train a model with just a couple of examples. Back in 2014, when AI started picking up steam, there were very few people that could apply it, today is the reverse. A high school kid can use it. So I don't think stricter regulation is going to stop EU from havi
    • by toutankh ( 1544253 ) on Thursday September 29, 2022 @12:31AM (#62923025)

      Let's imagine for a second that a self-driving car has a bug that causes an accident. Who, if not the car maker, should be held responsible?

      • by AmiMoJo ( 196126 )

        It depends how autonomous it is. Up to level 2, which is driving aids and a qualified driver carefully monitoring them, the manufacturer will always try to blame the person behind the wheel. Tesla does this all the time.

        For level 3 and up, where the driver can stop paying attention for at least some of the time, the manufacturer is entirely responsible during those periods. Sounding a warning for the driver to take over 0.3 seconds before impact doesn't absolve them either. I say that because when Tesla kil

    • You don't seem to grasp what the legislation actually does. The legislation introduces a risk liability. Which is paired with a lower than normal compensation for damages, because the liability is 'wider'. It requires importers and producers both to be able to prove their product wasn't defective, if a case comes up where it is plausible to question this in court. There are no 'American' punitive damages or very high sums involved in this, unless the issue exceeds the risk based liability and is plain a
    • by pat888 ( 2188172 )
      The EU is pretty competitive in aviation
    • Re: (Score:2, Troll)

      by AmiMoJo ( 196126 )

      The alternative is what you have in the US, where Tesla's beta test regularly kills people, and their surviving family members struggle to sue because Tesla just blames the driver and hides behind the EULA.

      We shouldn't sacrifice our right to redress, or our privacy, just because big businesses claim it will "kneecap" them. They said the same about GDPR, but it turns out that they can in fact comply with it just fine, and it didn't destroy the web in Europe.

      • We shouldn't sacrifice our right to redress, or our privacy, just because big businesses claim it will "kneecap" them. They said the same about GDPR, but it turns out that they can in fact comply with it just fine, and it didn't destroy the web in Europe.

        No, but constant cookie notifications on many sites is annoying.

        • by AmiMoJo ( 196126 )

          Cookie banners are annoying, but also non-compliant. None Of Your Business (NOYB) has filed complaints about hundreds of sites over that, and I've done a few myself. The regulators are a bit slow though.

          Hopefully in time cookie banners will go away. They violate Recital 32 of the GDPR, which states that to be complaint permission must be freely given. Any coercion like full screen banners that make the site impossible to use until dismissed, or making it harder to decline than to accept by putting the decli

      • Not unlike the convenience over security model in the software industry people are really keen on having the latest shiny thing to show they are better than everyone else. They only care when something happens to them and then it's "Why didn't someone do something before?"
      • And GDPR accomplished and added nothing useful whatsoever. My job was already about 50% InfoSec when GDPR rolled down on us. So I was pretty close to the "front lines," as the cliche goes. You know how much my job changed? You know how much my teammates' jobs changed? You know how much the directors of InfoSec or ProdOps jobs changed? Not at all. We were already following industry best-practices, and then some, to protect our company's data. To any of our knowledge; our efforts were already, and rem

        • by AmiMoJo ( 196126 )

          GDPR isn't really about security, it's about who owns your data. You own your own data, and companies can't just steal it.

  • > Under a "right of access to evidence", victims can ask a court to order companies and suppliers to provide information about high-risk AI systems so that they can identify the liable person and the fault that caused the damage.

    Lawyer: Your honour, the plaintiff identified the bug was the regularisation term in the loss function. They should have used L2 instead of L1, and used a different random seed; this was discovered using a non-causal inference method called "hunching" because the law doesn't r
  • by Anonymous Coward

    Notice how the governments always make laws to suite themselves way more than any citizens?

    This law can be used to wipe out individuals that use AI to protect themselves from the government. "Perky little shit, lets slap him with a law suite, see if he can stay afloat with that."

    In the Netherlands, the government made a law that says that "Anyone with a secrecy obligation cannot be found guilty of perjury for lying in court". They quoted "For example a lawyer that has a secrecy obligation to their client".

  • by bradley13 ( 1118935 ) on Thursday September 29, 2022 @03:38AM (#62923191) Homepage

    Why single out AI systems? If a product is defective, it is defective. Product liability laws already exist.

    The only thing special about AI is the fact that literally no one knows why an AI does something. But that also doesn't matter. If it has a provable error ("car thinks full moon is a stop light, stops in middle of road"), then it is defective. It doesn't matter whether that error is in a logical rule or buried in a neural net. Prove there is an error, get compensated for your damages.

    • When Teslas slammed into a concrete barrier or a firetruck, did they get sued the same as if the tires fell off the car?
  • What exactly is AI-enabled products? Do these idiots not realise that these are just software systems? There is nothing special about these systems, it's just software. Neural nets are basically just gigantic if-then-else lines of software.

    • What exactly is AI-enabled products? Do these idiots not realise that these are just software systems? There is nothing special about these systems, it's just software. Neural nets are basically just gigantic if-then-else lines of software.

      Now that, is what I call shortsighted. Not sure why you think laws written by and for Greed today, won't be still sitting in stone a half century from now, when those if-then-else systems are far more than that.

      These laws, are to protect those so they can develop full AI without restriction, regulation, or remorse. Of course, the irony of that is even the father of AI will be viewed as a pointless meatsack, and will be considered a mere annoyance by AI not unlike corporate fines are today.

  • So if an open source AI system is used that caused harm, will the person contributing the code be held liable? The person who selected an open source solution? Both?
    • The company releasing the system has the onus to certify that as safe and that all possible scenarios in the real world have been anticipated and tested.
      • that all possible scenarios in the real world have been anticipated and tested.

        Too bad the law didn't include the politician's solution to the Halting problem [wikipedia.org].....*facepalm*

  • Insurance rates in the EU will be insane for these vendors. And if you want more inflation, THIS is how you get more inflation.
    • Bottom line: this is a major threat to the EU's already waning ability to innovate. They are shooting themselves in the foot.

      Manufacturers will selectively rationalize, disable, or find substitutes for AI in the EU versions of their products. By "rationalize" I mean product teams will have to spin gears going back and forth with legal convincing each other that a particular application of AI isn't a risk (such as using AI during the design phase where output will be extensively human-reviewed) or the risk i

  • Wait ‘til the AI Attorneys hear about this!
  • ... is that "Artificial Intelligence" doesn't have any hard definition that makes it unambiguous when the law should actually apply. This means that it can be enforced at random, treating different people or different organizations differently.
  • The core issue here is that people want to pretend that AI is magic and cant discriminate or break laws; it honestly does all the time.

    We need greater transparency. and public integrity of these systems, and to do that, you have to be able to validate conclusively that they will not discriminate illicitly.

  • by fluffernutter ( 1411889 ) on Thursday September 29, 2022 @03:29PM (#62924899)
    As expected, Europeans are praising that governments are actually trying to protect the average person from behemoth corporations, and Americans are panicking about it.

Every nonzero finite dimensional inner product space has an orthonormal basis. It makes sense, when you don't think about it.

Working...