[Webinar] Red Teaming AI Agents: Find Vulnerabilities Before Attackers Do

​A live walkthrough of automated adversarial security testing with orq.ai

​AI agents deployed in production — with access to tools, memory, and user data — have a larger and less-tested attack surface than traditional software. Prompt injection, system prompt leakage, goal hijacking, and excessive agency are attack vectors that standard evals don't cover. The only way to find out if your agent is exposed is to attack it.

​In this webinar, the orq.ai research team walks through red teaming: what it is, why traditional testing misses agent-specific risks, and how to run automated adversarial testing against your own agents.

​What we'll cover

• ​What red teaming means for LLMs and AI agents — and how it differs from standard evals

• ​Live demo: running `evaluatorq redteam` against an orq agent — capability classification, attack generation, multi-turn execution, and final report

• ​How adaptive multi-turn attacks work — strategy planning, capability-aware targeting, and adversarial orchestration

• ​What the results mean and how to act on them: reading resistance rates and prioritizing fixes

• ​Industry frameworks (OWASP LLM Top 10, OWASP ASI) as a reference for what to test

​Who should attend

​AI engineers and platform teams shipping agents to production. Assumes familiarity with LLM agents; no prior security experience needed.

​Hosted by: Bauke Brenninkmeijer — AI Research Engineer

​Format: 45-minute session with live coding demo, followed by Q&A. Streamed live on Riverside