Prompted to Fail: When LLMs Go Rogue

LLMs are rewriting the rules of app security—and not always in a good way.

In this episode Alex sits down with Scott Rogers, a seasoned data scientist at ANvilogic to unpack why LLMs are the new wild west of application risk—and how old-school OWASP principles are making a serious comeback.

We cover:

• Real-world prompt injection failures (yes, including Air Canada’s rogue chatbot)
• How RAG systems can accidentally leak sensitive data
• Why GenAI risk ≠ traditional appsec—but it rhymes
• How classic tools like SAST, DAST, and logs can still save your bacon

Whether you're threat modeling your first LLM system or already knee-deep in GenAI, this episode is full of spicy detection ideas, war stories, and practical advice you won’t want to miss.

Stay in the loop! Connect with us:

• Join Dispatch Community: https://www.anvilogic.com/workshop
• Website: https://www.anvilogic.com/
• LinkedIn: https://www.linkedin.com/company/anvilogic 
• YouTube: https://www.youtube.com/@Anvilogic 

Detection Engineering Dispatch features candid conversations with security teams at top companies on how they build, measure, and scale world-class detection programs.