AI Can’t Detect Malicious Intent

Added: 2026-05-17

[00:00:00]That's the only difference. I mean, they're they're written with very often the same programming languages. They're, you know, in some cases they're generated by tools like, you know, cloud code and so on and so forth. But, it has no way of knowing the operator or the person that's asking it for something's intent. Is it with good intentions or is it is it with bad intentions? I mean, it can try and surmise. I mean, I did I I I can't remember. I asked it for something recently and it um it very much said no. And it said no and I it actually worried me. like, "Oh, are you going to remember this now? Are you going to hold this against me because I I asked it for some kind of an info." No, I asked Sorry, I asked it for an idiot killer. And it was like, "Hell, no. I'm not giving you an idiot killer."

[00:00:40]But, again, a couple of minutes later, I basically asked it for a mechanism whereby I could use W that policies to stop non-Microsoft binaries from being able to run.

[00:00:52]Now, that's effectively a W Sorry, that's effectively an idiot killer because it's going to stop non-Microsoft binaries from being able to run. And it gave it to me no problem at all. And again, it it's the difference between asking a question specifically for something bad and asking a question for something that ostensibly could be good, but is also bad.