When people think about the risks posed by genAI agents, they usually consider problems they'll detect immediately after the fact: agents deleting key files / databases / photos.

Another kind of agentic AI risk, and one that merits more attention, is the potential for the bot to create problems that are not immediately detected. Those can grow in the background until they become an existential threat.

If you've worked in finance, the hairs are already standing up on the back of your neck. Because you recognize this as the rogue trader threat. Barings Bank, SocGen, UBS, and others suffered billions of dollars in losses because traders' activities went uncovered for too long.

AI agents are poised to create the very same problems in every other industry vertical where they're deployed: