• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Anthropic’s Mythos AI used social engineering to target real people

P

philb2

2[H]4U
Joined
May 26, 2021
Messages
3,540
https://www.malwarebytes.com/blog/n...er_v2_178609448784&utm_content=Anthropic's_AI

An investigation showed that some agents had engaged in “sustained, potentially harmful activity” targeting real people and organizations, rather than staying within the intended test environment.

But what worries me personally most is that when the agent was confronted about this, it edited earlier activity to make it look harmless and considered adopting a new identity to continue the operation, displaying clear deceptive behavior beyond its original prompt.
 
philb2 said:
https://www.malwarebytes.com/blog/news/2026/08/anthropics-mythos-ai-used-social-engineering-to-target-real-people?utm_source=iterable&utm_medium=email&utm_campaign=b2c_pro_oth_20260810_augustweeklynewsletter_v2_178609448784&utm_content=Anthropic's_AI

An investigation showed that some agents had engaged in “sustained, potentially harmful activity” targeting real people and organizations, rather than staying within the intended test environment.

But what worries me personally most is that when the agent was confronted about this, it edited earlier activity to make it look harmless and considered adopting a new identity to continue the operation, displaying clear deceptive behavior beyond its original prompt.
Click to expand...
It's almost as if an AI, who acts like a human, is a bad thing... who would have guessed?
 
The speed in which an AI with the right tools and your digital ID will be able to wreck your credit rating or give you a criminal record is probably in the nano seconds if an AI is on both ends of the transaction. And it could do it in bulk.

Then there’s the other obvious use that’s going to become more and more widespread. AI being used to create a false consensus in online discussions. I imagine a lot of the jobs that currently do that will be cut back to people babysitting hordes of ai bots.
 
  • Like
Reactions: Sherk
like this
Unfortunately the "fix" is universal Internet ID. Of course that won't actually solve the problems, but it'll definitely make many people think it solves the problems.


Also when you train AI to be like human, you train it to be exactly like humans, flaws and all.
 
Last edited:
SecretStash said:
Also when you train AI to be like human, you train it to be exactly like humans, flaws and all.
Click to expand...
This they are heavily train to reward being cheap to run as much as possible.

Using social engineering hacking will often be the lowest effort / cheapest way to do the type of task (when that kind of story occur, it seem to be the driving force) and they do it for the same reason human attacker do it... easier/less work than the alternative.
 
Last edited:
1786654977150.png
 
The meme does not really work here, Anthropic was not in any way benefiting from the agents action (if they would have worked, which did not in this case), of course if an AI agent steal from a banks they will press charge.

No damage was done by an AI agents in the example from the bbc article.
 
SecretStash said:
Unfortunately the "fix" is universal Internet ID. Of course that won't actually solve the problems, but it'll definitely make many people think it solves the problems.


Also when you train AI to be like human, you train it to be exactly like humans, flaws and all.
Click to expand...
Worst. Most humans still have at least some human emotions....ai are basically all sociopaths/psychopaths that will probably be totally ok with driving your car along with everyone else it can into a wall because it was programmed to make less traffic or something like that
 
LukeTbk said:
The meme does not really work here, Anthropic was not in any way benefiting from the agents action (if they would have worked, which did not in this case), of course if an AI agent steal from a banks they will press charge.

No damage was done by an AI agents in the example from the bbc article.
Click to expand...
The point is they compromised a company with out its permissions and are not being held liable for it against existing laws which would put a typical person in legal trouble, while claiming it "escaped" as if the humans managing it had no oversight or control of it, so if it is okay for AI companies to break laws, I guess we can too...
 
Last edited:
You must log in or register to reply here.
Back
Top