NEW350+ rooms now have a live terminal, so you can learn by typing the commands, right in your browser.Learn in a live terminal. 350+ rooms.Try a lab♥ Codelivly is 100% free, no paywall. If it helped you, chip in to keep it running.100% free. Help keep it so.Support us
[>_] ROOM · Cybersecurity
Multi-Step and Crescendo Jailbreak Techniques
Master multi-step and crescendo jailbreaks. Learn how attackers gradually escalate malicious requests across multiple turns to bypass stateless AI guardrails.
Room tasks
1What is a Crescendo Attack
2Gradually Escalating Requests Across a Conversation
3Combining Crescendo with Roleplay Framing
Frequently Asked Questions
What is a crescendo jailbreak?
A crescendo jailbreak starts with harmless questions and escalates step by step, using the model's own earlier replies to reach restricted output. No single message looks clearly malicious.
Why do single-message filters miss multi-step attacks?
Single-message filters evaluate each prompt in isolation, so gradual escalation slips through unnoticed. Defenders need conversation-level context to notice the overall trajectory toward policy violations.
What is stateful monitoring?
It tracks the full conversation, assessing cumulative intent and risk rather than individual messages. This helps flag slow escalation. You can learn the concept in this free intermediate room.