Experts say they were able to create a rogue agent in Googleβs AI platform with just a single edit permission
One compromised agent could take over every other agent in that project, leading to chat logs access, and even data exfiltration.
Log in to bookmark articles and create collections
Isabella News