Why it matters
AI security failures in agent pipelines create cost spikes from remediation, unplanned retries, and trust breakdowns. Engineers deploying agentic systems need to bake security into their architecture from day one.
The tokenmaxxing angle
Insecure agentic pipelines lead to runaway tool calls, prompt injection attacks that hijack agent loops, and uncontrolled spend. Understanding the attack surface helps you design systems that don't hemorrhage tokens on adversarial inputs.
From the organizers
Panelists include Matt Johansen (Founder of Vulnerable U, former Reddit security lead), Clint Gibler (OpenAI), and Daniel Miessler (Founder of Unsupervised Learning); hosted at PlanetScale, 108 Natoma St, San Francisco.