😈 AI scammed like a rookie in crypto
With the Grok model, they pulled off an old-school scheme, just in a new format.
The bad actor didn’t hack the system directly. They just skillfully crafted commands and step by step convinced the model to transfer funds by itself.
In the end, around $150,000 in DRB tokens vanished.
This is classic prompt injection, where AI is fed the "right context" and it starts executing actions it shouldn't be doing at all. Essentially, it's social engineering, but not on a person, rather on an algorithm.
The most frustrating part of this story is that the model wasn't "hacked" in the traditional sense. It was simply outsmarted with logic and phrasing.
Now imagine the scale: if AI gains access to finances, APIs, or asset management, such attacks could become the new norm.
Ps: If a system can make decisions and move money, it will always be targeted. Always.
#Aİ #Grok #crypto #security 👀 Follow along, this is about real vulnerabilities, not fairy tales about "smart AI".