Comment by parkerrr

Being Aussie, I suspect given our response to most socially impacting issues is the guard rails will involve a ban on running and consuming AI services that don't meet some standard, with expensive licensing and certification. The effect will be only the big players will be allowed to operate, killing competition (side effect making home inference effectively illegal), no real accountability as "we are compliant", pre-defined slap on the wrist consequences for breaches, pushing all risk onto the public via acceptable use policy. Government is likely waiting for ISO to publish AI standards, so they can get one of the big 4 to recommend policies, which will then become law.

I suspect LLM weights will be treated like nuclear material, and there will be a thriving black market in AI models and services when all is said and done. Then our security and intelligence apparatus will align after identifying and shutting down rogue operators, and it will be the responsibility of everyone else to be secure against attack (already is for government entities, not that I disagree but shouldn't excuse borderline criminal behaviour).

But very little accountability for the big end of town. The laws already exist to pursue damages against companies whose systems breach others. They just need to be applied, but I suspect this is going to be the wedge to drive through a bunch of laws that protect big money and punish the little guy. I would like to be wrong.