Red Team

Thirty days in, on the last one, looking back over the dark arts covered along the way, with a few honest words for anyone just walking into AI security.

Someone asks you to go and see whether an AI system is safe. Where do you start? Map what it takes in and what it can reach, then turn the points you need to test into a checklist.

Hands on with garak, an automated AI red teaming tool.

Attacking AI means working out where the whole AI system can be misled, abused, or bypassed into doing something nobody designed it for.

One prompt can make an AI write code, organize documents, and even operate tools. So what happens when someone writes a bad prompt on purpose?