Where AI Breaks: Field Notes from GenAI and Agentic Testing
Bishop Fox consultants Derek Rush, Michael Cheng, and Katie Ritchie share what they're seeing in real GenAI assessments and LLM penetration tests. From recurring weak spots to agentic system failures, this session covers where AI security quietly breaks down and what to do about it.
AI is moving faster than most security programs can keep up with. Teams are shipping GenAI features, wiring up LLMs to internal data, and handing agents the keys to take real actions. The question that keeps surfacing is a simple one: what happens when someone decides to break it?
For the past year, Bishop Fox consultants have been answering that question in the field. They've assessed GenAI deployments, run penetration tests against LLM applications, and torn into the agentic systems teams are standing up right now. The findings are revealing, occasionally alarming, and genuinely useful if you're trying to build or defend any of this.
Watch Bishop Fox consultants Derek Rush, Michael Cheng, and Katie Ritchie as they pull back the curtain on what they're seeing. Together they'll cover:
- The recurring weak spots showing up across GenAI assessments
- What a real LLM penetration test report actually contains, so you know what rigor looks like before you commission one
- How to embed security into agents while they're still in development, when fixing things is cheap
- How to put agentic solutions to work on your own security data
- What architecture reviews of agentic implementations reveal about where these systems quietly fall apart
Whether you're building with AI or racing to secure what your teams have already shipped, you'll leave with a sharper sense of where the real risk lives and what to do about it.
