Post Content
Welcome back to AI Red Teaming 101!
In this episode, Gary Lopez, Principal Offensive AI Scientist on Microsoft’s ADAPT team, explores multi-turn attacks—adversarial techniques that steer models into bypassing safety protections. You’ll learn about two advanced approaches: Skeleton Key, which overrides safety instructions, and Crescendo, which incrementally guides models toward harmful outputs.
Gary walks through real-world examples and demonstrates how these attacks are executed in Microsoft’s red teaming labs, showing how even the most advanced models can be manipulated with persistent, well-structured inputs.
What You’ll Learn:
How multi-turn attacks like Skeleton Key and Crescendo work
Why gradual input manipulation can bypass model safety
How to simulate these attacks using Microsoft’s AI red teaming labs
✅ Chapters:
00:00 – Welcome & episode overview
00:20 – What are multi-turn attacks?
01:00 – Skeleton Key attack explained
02:00 – Crescendo attack strategy
04:00 – Real-world example: historical context to harmful output
05:30 – Demo: Crescendo in action
07:00 – Key takeaways & next steps
✅ Links & Resources:
AI Red Teaming 101 Episodes: aka.ms/airt101
AI Red Teaming 101 Labs & Tools: aka.ms/airtlabs
Microsoft AI Red Team Overview: aka.ms/airedteam
✅ Speakers:
Amanda Minnich – Principal Research Manager, Microsoft AI Red Team
LinkedIn: https://www.linkedin.com/in/amandajeanminnich/
Webpage: https://www.amandaminnich.info/
Gary Lopez – Principal Offensive AI Scientist, ADAPT
LinkedIn: https://www.linkedin.com/in/gary-lopez/
#AIRedTeam #AIRT #Microsoft #AI #AISecurity #AIRedTeaming #GenerativeAI #Cybersecurity #InfoSec #cybersecurityawareness #PromptInjection #CrescendoAttack Read More Microsoft Developer
