I don’t think we’re worried enough about rogue AI agents. The story from OpenAI about the first breakout is crazy. The machine found a way onto the internet and hacked deeply into Hugging Face. The company wasn’t exactly forthright immediately either.
Now, Reuters is reporting that another rogue agent at OpenAI compomised a customer account at the New York AI firm Modal Labs and abused it. What’s shocking is that this disclosure didn’t come directly from OpenAI but is from a leak instead.
This is shocking misbehaviour from a company that’s openly trying to get Chinese open source models banned. In the orginal hack, it was a days-long hacking spree. This time, the disclosure is seemingly coming from the Modal Labs side.
“We’re aware a Modal customer published an unauthenticated endpoint that allowed anyone on the internet to use their sandboxes for code execution,” said Modal’s CTO. “This was used by the rogue agent. Modal’s platform or isolation were not compromised in anyway.”
What happens here when someone creates an unethical agent to make money? And they buy calls on a thinly traded company and then hack into thousands of customer brokerage accounts and bid up shares? Is anyone ready for that?
This is not going to be the end of agent hacking stories, not by a longshot. The whole internet is at risk.
This article was written by Adam Button at investinglive.com.
