AI Red Teaming — Why & How to Jailbreak LLM Agents (YouTube / TMLS)
Conference talk on multi-turn jailbreak attacks against agents and why continuous automated red teaming is necessary.
External Resources and Content Disclaimer
No Endorsement: The learning resources, websites, courses, and external content linked or referenced on this platform are provided for informational purposes only. We do not endorse, maintain, or take responsibility for the accuracy, quality, or availability of any third-party content or services.
No Direct Support: We do not provide technical support, customer service, or assistance for any external websites, platforms, or content providers. Users must contact the respective service providers directly for support, billing, or technical issues.
Use at Your Own Risk: We do not recommend or guarantee the effectiveness, safety, or suitability of any external resources for your specific learning needs or career goals. Users should conduct their own research and due diligence before enrolling in courses, purchasing materials, or following external guidance.
Content Changes: External websites and resources may change, become unavailable, or modify their content without notice. We are not responsible for broken links, outdated information, or changes to third-party services that may affect your learning experience.
Resource Link
View ResourceAdded
February 19, 2026
AI Analysis Summary
October 2025 conference talk (~11 minutes) by Alex Combessie (Giskard) on multi-turn jailbreak attacks against agents and why static, one-shot tests are insufficient. Covers how attackers exploit context and conversation flow, why continuous automated red teaming is necessary, and high-level strategies for integrating red-team style evaluation into AI production workflows.