Microsoft identifies seven new ways AI agents can be hacked

The more you look at AI agents, the more it seems can go wrong with them, so Microsoft has extended its taxonomy of failure modes to help with discussion of possible fixes.

Raccontata dainfoworld.com