If a large web providers car kidnaps or maims you, your family should be entitled to multiple years of their sales across their entire portfolio. This begins to approach or accurately reflect the power of the company exerting its influence on society, political and legal systems in order to justify this beta test in public spaces. MVPs are for SaaS, not for impaling fleshbags like us with subpar code that works best in controlled environments. IMO It's the only fair trade for these companies testing these currently dangerous products unsupervised in your neighborhood without your express and explicit permission.
“But did you know our products will destroy all humanity?”
https://susam.net/inverse-laws-of-robotics.html
> Humans must remain fully responsible for decisions involving AI and accountable for the consequences arising from its use. If a negative outcome occurs as a result of following AI-generated advice or decisions, it is not sufficient to say, 'the AI told us to do it'. AI systems do not choose goals, deploy themselves or bear the costs of failure. Humans and organisations do. An AI system is a tool and like any other tool, responsibility for its use rests with the people who decide to rely on it.
eg. "oligarch" is used more frequently by more people
Well, it depends. If I'm using it according to the instructions and the tool does something it's not supposed to, the manufacturer is liable. If I take a weed whacker and start cutting fiber lines with it, I'm liable.
So in accordance with standard liability practice, AI labs ought to be liable for unintended actions their products take, and the user for illegal actions they direct the product to take.
The slight caveat is that a model just produces text. It can't hack anything on its own. I would say that the product in this scenario would be the harness/agent - if you build a system that takes untrusted input (LLM output) and takes potentially dangerous action based on it, well, that sounds dangerous? Possibly negligent?
I'd like to emphasize that it's not just "don't trust it when it says to rm -fr", but even a fully constrained boolean output can be negligently-trusted if it's hooked up to something impactful.
Like if any of us got caught doing what OpenAI apparently has done, we'd be in federal prison pending trial and hoping for a plea deal with some light at the end of the tunnel.
1) FTC chairman says AI agents will be treated just like any other tool (“resist anthropomorphizing”)
2) Studying enforcement of rules around personalized (“surveillance”) pricing, specifically rideshare, airplanes and food delivery
3) Public comment period open now on new rule to punish companies like Meta and Google that advertise scams.
The last one seems closest to happening but it’s still just a public comment period so it might die.