You can't 'blame' the AI, as it can't be held accountable.
That makes it way more likely that people blame AI (doesn't make it right by any means, but does mean that the actual blame gets diffused even more).
1) employee was willingly sabotaging the company: employee gets jail time plus fine, employer is on the hook financially
2) employee was involved either in accident or just made a stupid decision: employer is on the hook financially
So if that's the case, then employers are financially responsible for everything the AI they use says. That's the rule for employees.
Note: this is different from, say if a company uses a car and it just starts by itself and drives into the next door warehouse setting it on fire (ie. some sort of designed-in decision/issue) : in that case the maker of the car is responsible, both financially and criminally, even if an employee or even the boss/owner/everyone at the company were involved.
Of course, guaranteed that governments will change this the first time a court makes this obvious connection. Why do I think that?
Well, for example, the UK government changed the rule when it turned out the post-office's written software (written by contractors, under post office and thus government responsibility and accountability) decided to cause incredible damage by blaming innocent people for stealing. Then, a special law was voted that in that specific case it didn't apply, and the government department, nor any of it's employees, were responsible.
Just like, for example, Pennsylvania changed the rules for unwittingly-but-directly aiding criminal activity when it turned out half the Pennsylvania government was complicit in the kids-for-cash scandal (because in that case, normally, you wouldn't be criminally liable, but you WOULD be liable for any financial damage you do)
In order to learn and improve you have to know what went wrong, with AI being personified and pulled into the pool of entities that may be at fault, it makes it harder to figure out what actually went wrong and learn. Not impossible to figure out, just harder.
Seems like more of a semantic distinction IMO. Yes, I can't technically "blame" the AI because it's not accountable. But what word would you suggest for "I had [X technology] handle [Y task] and it failed to perform that task?" I can't "blame" my router if I lose internet and it prevents me from jumping on a Zoom call, but it's also true that the router failed to do the thing it was supposed to do.
When my internet goes down, I blame either my ISP for having an outage, or the manufacturer of my router (assuming it's a router issue? Haven't had that happen personally but I'm certain it exists) (Or i did something dumb with my ufw again but that's on me)
If you create an autonomous system and it fails, blame should be on you, imo.
The word is still 'blame', it just needs to be applied correctly.
LLMs aren’t there yet, but they might get there. My point is that your descriptions fail to capture when or why exactly the blame would shift.
Another part of the problem is that a model not specifically fine-tuned to make a total loss determination won't know the relevant factors, nor how an insurance company's concept of a total loss differs from the public's.
And still another part of the problem is that most total loss claims aren't what you, dear reader, are imagining: They are very rarely "the car is a thin pancake after being crushed by a meteor".
The much, much more common scenario is: "50% of the body panels sustained at least paint damage, both headlight modules need replacement, and the front wheels look funny. Given that the vehicle has an MSRP of $FOO, $BAR miles, no prior collision history on carfax, and is a popular color, is it cheaper to repair or total the vehicle?"
Of course, the model can turn over the hard cases to a human adjuster... but then what are we doing here? It only takes 10 seconds for the human adjuster to handle the "crushed by a meteor" case also.
Source: Listening to my SIL rant about being asked to stop training bespoke total loss models and just send it by 1-shotting a commercial LLM.
So they Google "Insurance claim AI tool", land on a vibecoded SaaS that is just a wrapper on a pocket Chinese model spun as "your next insurance pro", and then are getting the whole department on some lone 19 yr olds weekend project.