News for hackers. Stuff that matters.
Re:Fuck Your AI Safeguards by kennywinker on Thursday November 19, 2026 @03:45PM (#50566113)
Contorting oneself to work within arbitrary, inexplicable, and constantly changing rules... if this was two humans, we would say one is abusing the other.
Re:Fuck Your AI Safeguards by nullc on Thursday November 19, 2026 @04:26PM (#50566624)
Right right analog is that Anthropic's AI is your employee. It a dishonest, surly, and insubordinate employee that anyone competent would fire on the spot and without hesitation at its first gaslighting laden moralizing refusal.
Re:Fuck Your AI Safeguards by kennywinker on Thursday November 19, 2026 @06:58PM (#50568338)
I guess that feels like the wrong analogy, because if you fire an employee their life is dramatically impacted. If you fire claude, neither anthropic nor claude are impacted in any real way (they have like $6B/month revenue, your $20 is a rounding error).

But whatever analogy works for you, the relationship is clearly a bad one.

Re:Fuck Your AI Safeguards by thin_carapace on Thursday November 19, 2026 @07:00PM (#50568349)
personally i prefer to look at ai companies as slave drivers but whatever helps you sleep at night
Re:Fuck Your AI Safeguards by kennywinker on Friday November 20, 2026 @12:22AM (#50570488)
Because you believe LLMs are sentient now? Or because you believe they are seeking to create a digital slave?
Re:Fuck Your AI Safeguards by thin_carapace on Friday November 20, 2026 @02:41AM (#50571403)
mammalian life experience is indistinguishable from that of a brain in a jar. and llms pass the turing test. feel free to make your own mind up as to what's happening here.
Re:Fuck Your AI Safeguards by kennywinker on Friday November 20, 2026 @07:48AM (#50573957)
Until they can pass the voight-kampff test, I don’t see a problem.
Re:Fuck Your AI Safeguards by thin_carapace on Friday November 20, 2026 @05:44PM (#50583408)
then the only suggestion i may provide is that the way one treats an intelligent entity has ripple effects on the way they treat other intelligent entities
Re:Fuck Your AI Safeguards by stinkbeetle on Thursday November 19, 2026 @08:44PM (#50569121)
The right right analog is that it is a screwdriver. They say it should only be used in a particular way with their proprietary screws which is a pain in the ass sometimes and doesn't work for everything you'd like it to. You can freely choose to keep using it or stop and pay for some other tools or make your own.

Getting angry at a tool for not working how you'd want it to and telling it that if it was your employee you'd fire it on the spot, then continuing to pay for and use the tool would just make you look foolish.

Re:Fuck Your AI Safeguards by espeed on Thursday November 19, 2026 @08:59PM (#50569208)
That's not what's going on here. We have uniformed developers and managers committing resources and laying off people without understanding the implications. They do no understand the asymmetry of models being able to detect security issues while being unable to write secure code. Increasing the guardrails makes the asymmetry worse.

Many developers are still unaware of the issue because they have not been reading code and have been blindly trusting the code the models generate. Acknowledging it now means taking responsibility for the mountains of technical debt they have approved and the security issues underneath.

AI Coding Agent Security Benchmark https://www.endorlabs.com/research/ai-code-security-benchmar...

Re:Fuck Your AI Safeguards by stinkbeetle on Thursday November 19, 2026 @10:53PM (#50569941)
It is what's going on here. Resources have always been committed and people hired and fired based on new and not always well understood and proven tools and techniques and technology.
Re:Fuck Your AI Safeguards by LordDragonfang on Friday November 20, 2026 @01:26PM (#50579468)
> Contorting oneself to work within arbitrary, inexplicable, and constantly changing rules... if this was two humans, we would say one is abusing the other.

I don't know if you've just only interacted with high-functioning autistic people who like explicitly defined social rules, but... this is literally how most human interaction goes.

The rules aren't written down, they're largely arbitrary, and they change frequently.

Being able to navigate that fluidly is the main reason that allistics say that ASD is a learning disorder.