Field notes from the red team.

What actually breaks in deployed AI, why automated tools mislead, and why the people who built a system can never be the ones to test it. Written from real engagements, with the identifying details removed.

■ Foundations

What AI red teaming actually is, and why your pen test missed it

Australian organisations shipped AI faster than anyone tested whether it could be trusted. A penetration test checks the network and the API. It does not look at the model. Here is the gap, why it matters, and why a scanner will not close it for you.

Read the article →

If you have deployed AI that matters, I will tell you how I would try to break it.

Independent, onshore, authorised, evidence-sealed.

Start with a conversation