quote:
Leaked internal documents reveal Anthropic's "Claude Mythos" as a next-generation AI model in early testing, described as a major capability leap but posing "unprecedented cybersecurity risks" due to potential exploitation in automated attacks.

Be polite to your AI models. Please, thank you, may I, etc.

Don't say i didn't warn you.

0 ...

Report Post

Posted by grizzlylongcut

Member since Sep 2021

15371 posts

Posted on 3/27/26 at 8:49 am to The Mick

quote:
That's the scary part in all of it. The government has no one remotely close to the level of the tech lords. How can you try to regulate something you have no f'in clue how it's being done?

Simple, just say “yall are done. Pack your shite and get”

Problem is no other government would do it though so it really is a no going back situation.

0 ...

Report Post

Posted by UnitedFruitCompany

Bay Area

Member since Nov 2018

4081 posts

Posted on 3/27/26 at 8:54 am to Dire Wolf

quote:

I know a guy at Google that has been relatively apathetic about AI. At least compared to the twitter AI doomers

The other day at a school function he freaked the hell out on me about the scams they are capable of running and the volume with which they are going to be doing it

Basically anyone that handles any money at any company is going to flooded with increasingly sophisticated phishing well beyond the almost looks like the correct email

One of the problems is that all our shite has been leaked several times over but it’s an ocean of information too massive for hackers to really deal with in a targeted. Now AI can run thru it

If I'm reading this correctly, and I like to think that I am, what your google buddy is saying is that we went through all this trouble to create AI so it can turn into Smaug instead of Skynet? That actually tracks, 4th wave of war wise too, which is the scary part.

Basically, Smaugnet will steal our money, guard it in a dank cave, and watch the hilarity unfold. Talk about a let down.

0 ...

Report Post

Posted by diat150

Louisiana

Member since Jun 2005

47776 posts

Posted on 3/27/26 at 9:00 am to MrLSU

quote:
Leaked internal documents reveal Anthropic's "Claude Mythos" as a next-generation AI model in early testing, described as a major capability leap but posing "unprecedented cybersecurity risks" due to potential exploitation in automated attacks.

this is a good thing. our enemies will use this against us... so we need to be able to do the same against them. we also need the same capabilities to watch and protect our infrastructure. anyone not embracing this is a fool. There is no putting the cat back in the bag.

1 ...

Report Post