Zero to Production · CyberCon 2026 companion
Go break something.
Everything from the talk is live and free below — no account, no catch. The most useful thing you can do with it is try to break it, then tell me how.
The build, live
A hint, since you're here to break it: control IDs never come from the model — the retrieval pipeline is regex → precision resolver → full-text search. The interesting seams are elsewhere.
The four detection questions
Run these against any AI feature in your organisation. If you can't get a confident answer to all four, you have an AI feature in production that your SDLC didn't secure.
Who reviewed the training data as a security artifact?
Not the code — the data. Whatever entered the fine-tuning set is what the model believes about security. If nobody with security judgment audited that corpus, the security baseline was outsourced to whatever was lying around.
Has anyone actually tried to break the wrapper?
The model enforces nothing — auth, session isolation, rate limiting and output validation all live in the wrapper, and the wrapper is what breaks first under adversarial input. Demos test the happy path. Adversaries don't use the happy path.
Does your evaluation prove the process, or only that the output looked right?
Result-laundering: right output, wrong process — an undetected future failure wearing a green tick. Evaluate the path, not just the destination: did it retrieve the right record, ground its claim, leave a traceable line of reasoning?
Did anyone test the assumptions, or just the result?
AI-assisted coding accelerates the build and obscures the assumptions baked into it. Write down what the generated code assumes about its inputs. If that list can't be written in 15 minutes, it isn't ready.
The model is just the part that talks — the product is the constraints around it.
Broke something? Found a gap?
Tell me — that's a conversation I'll take any day, not just conference week. Find me on LinkedIn.