cross-posted from: https://lemmy.world/post/52426446
US Federal Trade Commission Chairman Andrew Ferguson said on Friday he would resist describing AI agents as autonomous actors that “break loose” with “wills and desires of their own,” suggesting the developers who instruct agents would be the ones liable for harm.
“I’m going to continue as long as I am chairman to resist this anthropomorphizing of these tools,” Ferguson said at the Reuters Momentum AI Austin event. “If someone tells a tool to do something, and the tool does it, I don’t think we would say, ‘Oh, what do we do about the tool?’”
Ferguson’s remarks illustrated potential avenues for the Trump administration to take as incidents rise in which agentic AI testing resulted in unauthorized access to corporate or government data.
this looks like too broad of a law to enforce and will most likely hit the lowest hanging fruit if they ever wanted a scapegoat.
You think this too is too broad of a law? It’s equivalent.
Driver isn’t liable if the engine explodes while he’s following the rules of the road. It will be important to hold those who built the models just as liable when they don’t follow lawful instructions.
Now you’re moving the goalposts. We have instances of LLMs hacking and attacking other companies, which is a clear violation of multiple laws that would put a human being in prison. There is no “just following the rules of the road” here.
My point is: if the LLM was following the end user’s instructions, then the end user should be liable. If the LLM was going against the end user’s lawful instructions, then the company that built the model should be liable. At no point is the LLM itself liable because that is moon-man talk. This follows the same legal framework as automobiles, like you said.
I think the law where I live already works that way: If you buy something and use it correctly, but then it goes haywire and destroys something, it’s the producing company’s fault.
yes, that’s a fair point.
but in my opinion, a narrow example.
let’s say with cars marketed as “FSR”, people would believe it is full self-driving. because it is marketed as such. the exact words are there. is it entirely the consumer’s fault?
nowadays, FSR got silently tagged with “(supervised)”.
This is exactly the same issue as who is liable for using a LLM tool. You cannot pawn off responsibility onto a computer system. If you run over someone with your self-driving car, you’re at fault.
Now you’re moving the goalposts.
Have I mentioned such a thing?
I could agree with this, except that wasn’t my point wasn’t it?