- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
Leaked documents from Anthropic show new generation of AI which pose cybersecurity risks
Posted on 3/26/26 at 11:53 pm
Posted on 3/26/26 at 11:53 pm
Loading Twitter/X Embed...
If tweet fails to load, click here. 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.
The exposure occurred via a misconfigured public data cache containing nearly 3,000 files, including draft materials and plans for an invite-only CEO summit; Anthropic attributes it to a configuration error and has since secured access.
This incident echoes recent Anthropic vulnerabilities, such as a zero-click prompt injection flaw in the Claude Chrome extension affecting millions, underscoring escalating security challenges in frontier AI development.
12
Posted on 3/27/26 at 12:00 am to MrLSU
Posted on 3/27/26 at 6:21 am to okietiger
Posted on 3/27/26 at 6:58 am to MrLSU
They need to rein this in before something massive happens.
This post was edited on 3/27/26 at 7:01 am
Posted on 3/27/26 at 7:03 am to zippyputt
Who will do that? We don’t have a functioning government.
Posted on 3/27/26 at 7:04 am to MrLSU
Hurtling towards the Great Filter at breakneck speed.
Posted on 3/27/26 at 7:04 am to MrLSU
"leaked documents". Anthropic always putting out the doomer stuff every few months to keep people invested
Posted on 3/27/26 at 7:04 am to MrLSU
quote:
This incident echoes recent Anthropic vulnerabilities, such as a zero-click prompt injection flaw in the Claude Chrome extension affecting millions, underscoring escalating security challenges in frontier AI development.
I know this should be disconcerting, but really have no idea what any of that means.
Posted on 3/27/26 at 7:06 am to zippyputt
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?
They need to rein this in before something massive happens.
Posted on 3/27/26 at 7:09 am to MrLSU
quote:
such as a zero-click prompt injection flaw in the Claude Chrome extension affecting millions, underscoring escalating security challenges in frontier AI development.
Maybe laying off actual programmers who can review code is a bad idea and/or this is another ploy from "Security" firms who are willing to sell a product to combat this from happening.
This post was edited on 3/27/26 at 7:10 am
Posted on 3/27/26 at 7:10 am to NIH
quote:
Who will do that? We don’t have a functioning government.
Tech Bros financed this administration. There won’t be any federal oversight of AI until 2029 at the earliest. It’ll be too late by then.
We voted for this and have no one to blame but ourselves.
Posted on 3/27/26 at 7:13 am to Darth_Vader
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 of 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 to massive for hackers to really deal with in a targeted. Now AI can run thru it
The other day at a school function he freaked the hell out of 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 to massive for hackers to really deal with in a targeted. Now AI can run thru it
This post was edited on 3/27/26 at 7:15 am
Posted on 3/27/26 at 7:17 am to Dire Wolf
At least nostalgia will hit hard once shite hits the fan and we revert back to a world without internet.
Posted on 3/27/26 at 7:27 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 of 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 to massive for hackers to really deal with in a targeted. Now AI can run thru it
Reminds me of this situation a few months back that I ran into with a phishing email.
quote:
My standard response is that I will send a separate email to the (hacked) person to advise them they were hacked. Not even to test them, but to alert them (I'd hope people do the same for me).
YESTERDAY, when I sent the email, it sent one back. It responded to my separate email with, "No, I did send that document".
So phishing AI has gotten to the point where they will respond go separate email chains from the phishing attack to validate the phishing email. It's bad enough for normal people not to click these emails, but having the AI be able to communicate with you to falsely validate them is going to be a problem...especially if it can read the hacked person's emails to mimic their language/syntax style.
Posted on 3/27/26 at 7:35 am to MrLSU
AI will think we're retarded and just steal some rockets to go mine the moon and build solar panels in space.
No difference than humans ignoring dolphins and going about our day.
No difference than humans ignoring dolphins and going about our day.
Posted on 3/27/26 at 7:55 am to RougeDawg
I think AI is the OP just to throw us all off the trail. Hell, all you posters may be AI too. Maybe it’s just me on here…
Posted on 3/27/26 at 8:04 am to MrLSU
It's quite an interesting marketing strategy for sure.
"These catastrophic events will be made possible by AI. And you're all going to be fired. But give us money so we can continue at breakneck pace."
"These catastrophic events will be made possible by AI. And you're all going to be fired. But give us money so we can continue at breakneck pace."
Posted on 3/27/26 at 8:05 am to UltimaParadox
quote:
"leaked documents". Anthropic always putting out the doomer stuff every few months to keep people invested
More like every month to distract from the fact that they need to jack their prices up so they can turn a profit:
Loading Twitter/X Embed...
If tweet fails to load, click here.
Posted on 3/27/26 at 8:14 am to okietiger
It's coming
Posted on 3/27/26 at 8:38 am to jdd48
quote:
"These catastrophic events will be made possible by AI. And you're all going to be fired. But give us money so we can continue at breakneck pace."
China is going as far as smuggling certain AI chips into their country in order to best us in this new-aged space race.
Page 1 of 2
Popular
Back to top
Follow TigerDroppings for LSU Football News