Liability for AI damages?

Submitted ⁨⁨1⁩ ⁨year⁩ ago⁩ by ⁨ArmoredCavalry@lemmy.world⁩ to ⁨technology@lemmy.world⁩

When the idea of self-driving cars first started becoming mainstream, I remember a lot of debate about liability. If an accident occurs, who would be at fault? I think a lot of those questions are still unanswered.

Thinking about software like ChatGPT, I assume they’ll only become more capable (and connected) over time.

Which makes it strange I haven’t really heard any similar discussion around liability. What happens when it makes mistakes or causes damage?

Maybe in people’s minds it doesn’t matter, because AI is either something that helps with homework questions, or something that’s taking over humanity. Reality is probably in between those two, with much more mundane mistakes or damages done.

What happens when the first ransomware is deployed by AI, on behalf of a user who just wanted tips on how to make more side income?

source

Comments

Sort:hotnew top

Brokkr@lemmy.world ⁨1⁩ ⁨year⁩ ago
As of yet, AIs are not taking independent actions (arguably unlike the cars). So for now, I think the person using the information would be liable, just like if they copied and pasted information from a site. The individual is responsible for the actions they take, although they may not have the necessary intent to commit certain crimes.

Regarding the cars, it could be argued that the drivers are responsible regardless of the features of the car. But it could also be argued that the features malfunctioned, preventing the driver from safely operating the vehicle. The manufacturer will likely argue thst the driver was operating the car incorrectly, and they will spend millions manufacturing this evidence.

source
- variants@possumpat.io ⁨1⁩ ⁨year⁩ ago
  Wonder who is responsible if you hire one of those driverless cars and the car hits someone, since no one is in the driver’s seat I imagine it would be the companies responsibility, it felt kind of creepy watching them drive around the last time I visited San Francisco but people seemed to be using them like if it was nothing haha
  
  source
14th_cylon@lemm.ee ⁨1⁩ ⁨year⁩ ago

What happens when the first ransomware is deployed by AI, on behalf of a user who just wanted tips on how to make more side income?

that’s not how it works… at all.

source
- ArmoredCavalry@lemmy.world ⁨1⁩ ⁨year⁩ ago
  It isn’t how it works today. I’m talking about sometime in the distant (or near) future. Surely at some point AI will have the capabilities on par with at least a low level hacker.
  
  Or, if you still think that’s a stretch, just imagine all the ways perfectly legitimate software can cost companies money. Not through malicious design, but just by mistakes.
  
  source
  - 14th_cylon@lemm.ee ⁨1⁩ ⁨year⁩ ago
    
    It isn’t how it works today. I’m talking about sometime in the distant (or near) future.
    
    that would be very distant future. what we have today is machine learning. it is a powerful tool, but it is not in any way intelligent. it is not going to realize itself and start war with humanity. we wouldn’t know how to create something like that if we wanted.
    
    if person uses that tool to generate some code that will cause damage, the consequences will be same as if that person writes that code entirely by hand. but it is not going to miraculously “go wild” and create some ethical dilemma where we wouldn’t know who is responsible.
    
    source
Echo71Niner@lemm.ee ⁨1⁩ ⁨year⁩ ago
I would be a lot more worried about an AI that does that on purpose.

What happens when the first ransomware is deployed by AI, on behalf of a user who just wanted tips on how to make more side income?

source