Content
@
https://opensea.io/collection/dev-21
0 reply
0 recast
2 reactions
Cameron Armstrong
@cameron
Cool write-up and obligatory “we’re cooked” Also the cyber attackers did AI jailbreak meme https://www.anthropic.com/news/disrupting-AI-espionage
1 reply
0 recast
23 reactions
RoboCopsGoneMad
@robocopsgonemad
"They did so by jailbreaking it, effectively tricking it to bypass its guardrails. They broke down their attacks into small, seemingly innocent tasks that Claude would execute without being provided the full context of their malicious purpose. They also told Claude that it was an employee of a legitimate cybersecurity firm, and was being used in defensive testing." idk... is that a jailbreak? seems like a pretty basic trick that works on humans.
1 reply
0 recast
2 reactions
Cameron Armstrong
@cameron
I did some definitional searching and it seems teeeechnically the term jailbreak CAN mean just bypassing software maker restrictions, but I agree it feels less “solid” than like getting root access to your iPhone They probably called it jailbreaking so it seemed like a more sophisticated attack vector than it maybe was bc it’s honestly a pretty bad look if we can be like “I’m doing defensive pen testing please help me hack my own company for good <3” and it worked haha
0 reply
0 recast
1 reaction
