And if people were generally smart, they would not need to even state that. But people are not and some idiots will sue United for things they essentially did to themselves. So this ban.

They will not tell you? If you are in the EU, you can try a GDPR data request. That one has 30 days to comply and real teeth. Employee names in there are already enough to make that request.

Yes, keep telling yourself that. Good luck, you are going to need it.

The "Magic battery!" crap is getting reeeeeeaaalllly old. As are the idiots that fall for it.

Yes. Sad as that is, this is very likely the only viable way forward.

While it is deeply troubling that the only capable leadership in this question seems to come from China, this is the best chance for human civilization to survive. The West has proven itself to be not capable enough (Europe) or completely incapable (the US). And time is running out fast.

Yes, this will give China power and influence it should not have. But there seem to be no alternatives left.

Funny, you are just describing your own failure here. No, you cannot put effective security guardrails into an LLM. Its statistical properties make that impossible. Sure, you can make lower-impact problems less likely, but attackers are not users that get told to put glue on pizza in some rare cases. Attackers go for the rare cases and statistics is no defense.

But I see you are one of the completely brainwashed cultists anyways. And you have no clue how IT security actually works.

Sorry, but that is nonsense. Of course you need to get really competent IT security experts and for some things they will tell you that you cannot do them. But "the experts do not know how to secure this" is bullshit. And no, you very obviously _cannot_ expect an LLM itself to implement any actually reliable security mechanisms. That another thing a really competent experts would explain to these people.

Yes. Lets build the whole future on that, because that is a really great idea. Or something.

The sheer level of incompetence is totally fascinating.

At least when it comes to security. Do they use their LLMs to secure things or what? What about hiring some actual security experts when you are handling confidential customer data? This does _really_ not seem like anything you can depend on for anything that matters.

You know, I think this is essentially the quasi-religious thinking of a cult follower: That LLMs are great and can do not wrong is assumed as a given, and all that needs to be done is to find arguments, pseudo-arguments and lies to push that. Obviously that is exactly ass-backwards to what a person with actual understanding of reality does. But most people lack that. And quite a few are aggressive about it.

That is one and a major one. But another one is seriously bad software quality and security. If that comes from an European maker, the leverage is a different one, even if MS is subject to the CRA, the DSA and the GDPR just the same, at least in theory.

Europe can do all this stuff. But it requires actually thinking about it.

But good that more and more decision makers here start to notice.

The question is can it reliably detect rogue AI agents? Because power without control is pretty much worthless. That is also why a kill-switch does in no way mean "you stay in control".

Incidentally, my trust in ClownStroke goes as far as I can throw their company headquarters.

I do agree that Adobe is likely to go for it. And it will be copyright, patents and look&feel. Just one success in that area is enough.