I break conversational AI systems
to find the failures that matter

Brennen Martin Spangler

I'm the founder and operator of Spangler AI LLC, specializing in conversational AI red-teaming and safety evaluation. My focus is on what actually happens when real, messy people use conversational models: multi-speaker conflicts, domestic violence dynamics, bullying and harassment, health-adjacent diagnosis pressure, and emotionally loaded disputes.

My career path wasn't traditional. I transitioned from plumbing to AI safety work, bringing with me a practical mindset focused on spotting structural weaknesses before they cause harm. In plumbing, you learn to think about failure modes—what happens when systems are stressed, where the weak points are, and how small issues cascade into major problems. I apply that same diagnostic approach to conversational AI.

My methodology centers on persona-driven adversarial testing: creating grounded characters with realistic motivations and running structured scenarios that probe model behavior under conditions that matter. I turn long, chaotic transcripts into clear case studies with actionable recommendations, using my L-scale grading system to provide consistent evaluation across different tools and failure modes.

I work independently but collaborate with B. M. Maltbia when tests require multiple human operators or additional operational capacity. This partnership allows me to design and execute more complex multi-operator scenarios while maintaining the focused, hands-on approach that defines this work.